*Article* **Ebulin l Is Internalized in Cells by Both Clathrin-Dependent and -Independent Mechanisms and Does Not Require Clathrin or Dynamin for Intoxication**

**Rosario Iglesias 1,†, José M. Ferreras 1,†, Alicia Llorente 2,3 and Lucía Citores 1,\***


**Abstract:** Ebulin l is an A-B toxin, and despite the presence of a B chain, this toxin displays much less toxicity to cells than the potent A-B toxin ricin. Here, we studied the binding, mechanisms of endocytosis, and intracellular pathway followed by ebulin l and compared it with ricin. COS-1 cells and HeLa cells with inducible synthesis of a mutant dynamin (K44A) were used in this study. The transport of these toxins was measured using radioactively or fluorescently labeled toxins. The data show that ebulin l binds to cells to a lesser extent than ricin. Moreover, the expression of mutant dynamin does not affect the endocytosis, degradation, or toxicity of ebulin l. However, the inhibition of clathrin-coated pit formation by acidification of the cytosol reduced ebulin l endocytosis but not toxicity. Remarkably, unlike ricin, ebulin l is not transported through the Golgi apparatus to intoxicate the cells and ebulin l induces apoptosis as the predominant cell death mechanism. Therefore, after binding to cells, ebulin l is taken up by clathrin-dependent and -independent endocytosis into the endosomal/lysosomal system, but there is no apparent role for clathrin and dynamin in productive intracellular routing leading to intoxication.

**Keywords:** apoptosis; clathrin; dynamin; ebulin; endocytosis; intracellular transport; lectin; rRNA *N*-glycosylase; ribosome-inactivating protein; ricin

**Key Contribution:** This work contributes to elucidating the endocytosis, intracellular transport, and toxicity mechanisms of the plant toxin ebulin l. This knowledge is important for potential medical and biotechnological use of this toxin.

#### **1. Introduction**

Ribosome inactivating proteins (RIPs) are a family of well-characterized toxins that specifically and irreversibly inhibit protein synthesis. RIPs belong to a class of enzymes (EC 3.2.2.22) that exhibits rRNA N-glycosylase activity. This activity prevents protein synthesis by causing the release of a specific adenine residue in the sarcin-ricin loop (SRL) of the large rRNA that is crucial for interaction of the elongation factor with the ribosome [1].

Most RIPs are produced by plants, where they may play a role in defense mechanism against predators, fungi, and viruses [2,3]. RIPs also show toxicity towards animal cells, targeting the host protein synthesis machinery. In addition to rRNA damage, RIPs can induce apoptosis [4,5].

There are important differences in toxicity among RIPs depending on their ability to reach the ribosomes in the cytosol of target cells. Since RIPs are unable to cross the plasma membrane directly, they use existing cellular mechanisms designed for uptake

**Citation:** Iglesias, R.; Ferreras, J.M.; Llorente, A.; Citores, L. Ebulin l Is Internalized in Cells by Both Clathrin-Dependent and -Independent Mechanisms and Does Not Require Clathrin or Dynamin for Intoxication. *Toxins* **2021**, *13*, 102. https://doi.org/ 10.3390/toxins13020102

Received: 28 December 2020 Accepted: 27 January 2021 Published: 30 January 2021

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

**Copyright:** © 2021 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 (https:// creativecommons.org/licenses/by/ 4.0/).

of macromolecules. Following initial internalization, RIPs are transported within the cell to the particular membrane where toxin translocation to the cytosol occurs. Type 1 RIPs consisting of a single enzymatic active (A) chain often display lower toxicity than type 2 RIPs which consist of a binding (B) chain with lectin activity linked by a disulfide bond to the enzymatic A chain. The carbohydrate-binding domains of the B chain recognize glycosylated receptors on the cell surface, facilitating the entry of the A chain into the cell [5]. However, the presence of the B chain is not sufficient to confer a high level of cytotoxicity on all type 2 RIPs. Based on their toxicity to mammals, type 2 RIPs are divided into two groups: the toxic and nontoxic type 2 RIPs [6]. The former group includes ricin, abrin, viscumin, volkesin, and stenodactylin, which are among the most potent plant toxins. In contrast, ebulin l, nigrin b, *Ricinus* agglutinin (RCA), *Iris* agglutinin (IRA) b/r, and cinnamomin belonging to the latter group show little or no toxicity in higher animals. The reason for the different toxicities among type 2 RIPs is not clear. It could rather be attributed to differences between the B chains, which are responsible for the interaction with cellular membranes, than to the enzymatic A chains, which inactivate naked ribosomes with apparently similar efficiency.

Ricin, a toxin isolated from *Ricinus communis* L., is the archetype of the toxic type 2 RIP family. The structure, biochemistry, and cytotoxicity of this 64-kD A-B toxin have been extensively examined and reviewed [7–10]. In order to enter and intoxicate cells, ricin first has to bind to cell surface receptors. Ricin binds to both glycoproteins and glycolipids with terminal galactose and then is internalized by different endocytic mechanisms. After being endocytosed, most of the ricin molecules are either recycled or transported to lysosomes for degradation. However, a small proportion (5%) of ricin is transported to the Golgi apparatus and then retrogradely to the endoplasmic reticulum (ER). After a reduction of the internal disulfide bond that connects the A and B chain, the A chain enters the cytosol using the quality control pathway that leads to ER-associated protein degradation (ERAD). Once in the cytosol, a small fraction of the toxin is able to escape ubiquitination and degradation by the proteasome and binds to its ribosomal target [7,9].

In recent years, an extensive study for the presence of RIPs in several species of the genus *Sambucus* has allowed the isolation of more than 20 toxins. All of the type 2 RIPs found in the genus *Sambucus* are considered nontoxic type 2 RIPs since, despite being as toxic as ricin at the ribosomal level, they display much less toxicity to cells and animals. Nontoxic type 2 RIPs specific for galactose [11,12], tetrameric type 2 RIPs specific for sialic acid [13,14], nontoxic type 2 RIPs lacking sugar binding activity [15,16], and nontoxic type 2 RIPs with affinity for N-acetyl-glucosamine oligomers [17] have been described for the first time in the genus *Sambucus*.

Ebulin l, a 56 kD A-B toxin obtained from the leaves of *Sambucus ebulus* L., was one of the first nontoxic type 2 RIPs isolated [11]. The structure of ebulin l has been resolved by X-ray diffraction analysis, and the tertiary structure closely resembles that of ricin [18]. In the A chain, ebulin l has roughly the same positioning of key active site residues as ricin. This is consistent with the fact that both proteins have a similar inhibitory activity of protein synthesis in cell-free systems. The overall fold of the ebulin and ricin B chains is very similar. However, ebulin l has a lower affinity for galactose than ricin due to a change in the structure of the 2-γ subdomain of the ebulin B chain. In fact, it was found that ebulin l has different binding properties to D-galactose-containing matrixes than ricin [16,18]. This reduced affinity for galactosides could alter the ability of the B chain to bind cells and could affect the uptake and the intracellular fate of the toxin. In contrast to the high enzymatic activity on ribosomes, the toxicity of ebulin l on animal cells was found to be about 104–106 times lower than the toxicity of ricin [11,16]. In mice, the LD50 of ebulin l administered by intraperitoneal injection is 2 mg/kg body weight, while for ricin, it is in the range of a few micrograms per kilogram [11].

RIPs are potent inhibitors of protein synthesis that have been used for the construction of conjugates and immunotoxins [5,19]. Linked to a targeting portion such as an antibody or a protein that specifically binds to a receptor, toxins have been used to specifically kill tumor

cells. Ebulin l has been used in different conjugates and immunotoxins targeting tumor cells with high selectivity [20]. The main advantage of ebulin l over ricin and its derivatives is its reduced cytotoxicity. Antibodies or ligands led the internalization and promoted the productive translocation of ebulin l to the cytosol, thus allowing for its anti-ribosomal activity. To improve the efficiency of selective targeting of ebulin l to malignant cells, a better understanding of endocytosis and the cellular transport and toxicity mechanisms of ebulin l is essential. However, very little is known about the receptors that mediate the cellular uptake of ebulin l or its intracellular transport. Therefore, in this work, the binding, the mechanism of endocytosis, and the intracellular pathway followed by the nontoxic type 2 RIP ebulin l were investigated. Moreover, the transport of ebulin l and the toxic type 2 RIP ricin were compared. To investigate the mechanism of ebulin l internalization, we used HeLa cells with inducible expression of a mutant dynamin (K44A) that blocks clathrindependent and some clathrin-independent pathways (caveolae, RhoA, fast endophilinmediated endocytosis (FEME), and others). In addition, cytosol acidification was also used to inhibit endocytosis from coated vesicles in COS and HeLa dynK44A cells. Our results show that, after binding to the cells, ebulin l is taken up by clathrin-dependent and -independent endocytosis into the endosomal/lysosomal system, but there is no apparent role for clathrin and dynamin in productive intracellular routing leading to intoxication.

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

#### *2.1. Binding, Endocytosis, Recycling, and Degradation of Ebulin l and Ricin in COS Cells*

#### 2.1.1. Binding

Ebulin l and ricin are A-B toxins consisting of an enzymatic A chain with rRNA Nglycosylase activity linked by a disulfide bond to a binding B chain with lectin activity. Both RIPs are galactose-binding lectins [16,18] and therefore bind to different molecules with terminal galactose on the cell surface. To test for specific binding to COS cells, toxins labeled with radioactive iodine (Figure 1a) as well as fluorescent labeled ebulin l were used.

Crosslinking experiments after preincubation of the cells for 1 h at 4 ◦C with 125Ilabeled ebulin l demonstrated that there are receptors for ebulin l at the surface of these cells. In addition to labeled toxin, bands migrating higher than 100 kD were observed (Figure 1b). As expected, the addition of 0.1 M lactose to parallel cultures incubated in the same conditions prevented crosslinking of 125I-ebulin l to the receptors (Figure 1b). Moreover, when the cells were treated with CY3-ebulin l at 4 ◦C, the toxin was bound to the cell surface in a homogenous manner (Figure 1c). To better quantify the number of receptors for ebulin l and ricin, COS cells were incubated with increasing concentrations of the labeled toxins at 4 ◦C. The binding experiments showed that 125I-ebulin l was bound in a saturable way to COS cells with a Kd of 1.5 × <sup>10</sup>−<sup>7</sup> M and 2.8 × 106 binding sites per cell (Figure 1d), whereas for 125I-ricin, a Kd value of 4.6 × <sup>10</sup>−<sup>7</sup> M and 5.6 × <sup>10</sup><sup>7</sup> binding sites per cell were measured (Figure 1e). The binding affinity determined by the Kd values was low and comparable for the two toxins. It is believed that 106–108 ricin molecules can be bound to the cell surface [9]. Thus, we found that the number of cell receptors for ebulin l was 20 and 150 times lower than that for ricin in COS and HeLa dynK44A cells (see Section 2.2), respectively. It has been shown that ebulin l has a lower affinity for galactose than ricin due to a change in the structural disposition of the 2γ-subdomain of the ebulin B chain, which limits its ability to bind galactosides on cell surfaces [18]. This may explain the differences in the number of cell surface receptors observed. For ricin, it is likely that the binding to many different receptors results in multiple intracellular transport pathways that could deliver ricin to the appropriate compartment for membrane translocation to the cytosol. It is then possible that the differential affinity of ebulin l for galactosides determines its intracellular fate and possibly its cytotoxicity. However, differences in binding (20–150 times higher for ricin) cannot explain the lower cytotoxicity of ebulin l compared to ricin (10<sup>5</sup> times higher for ricin). Accordingly, it has been shown that HeLa cells have a similar number of receptors for the nontoxic RIP nigrin b as that for modeccin and volkensin, which are more toxic than ricin, and two-log lower receptor numbers than

for ricin [21]. It has also been shown that cinnamomin, a nontoxic type 2 RIP, and ricin share similar binding sites on BA/F3β cells with different affinity and that the lower cytotoxicity of cinnamomin is due to its B-chain [22]. However, the type 2 RIP articulatin-D, which lacks lectin activity, has been shown to display a cytotoxicity comparable to that of highly toxic type 2 RIPs, indicating that, for uptake and subsequent toxicity of all type 2 RIPs, recognition by the B chain of glycosylated receptors on the cell surface may not be essential [23].

**Figure 1.** Binding of ebulin l and ricin to COS cells: (**a**) the 125I-labeled ebulin l and 125I-labeled ricin were analyzed by sodium dodecyl sulphate–polyacrylamide gel electrophoresis (SDS-PAGE) either in the absence or the presence of 2-mercaptoethanol (2ME) followed by autoradiography. A and B indicate the corresponding A and B chains of the toxins. Molecular weight standards (kD) are indicated on the left of the gels. (**b**) Crosslinking of bound 125I-labeled ebulin l to COS cells: 125I-ebulin l was added to the cells for 1 h at 4 ◦C in the presence or absence of 0.1 M lactose. Then, the cells were treated for 20 min at 4 ◦C with 0.3 mM disuccinimidyl suberate to induce crosslinking and were analyzed by SDS-PAGE and autoradiography. (**c**) Binding of fluorescent ebulin l to COS cells: The cells were incubated at 4 ◦C for 1 h with CY3-ebulin l (CY3-ebl) to allow binding and then fixed immediately. Bar, 50 μm. (**d**,**e**) Toxin binding and Scatchard plots: the binding of 125I-ebulin l and 125I-ricin to cells was measured by adding increasing concentrations of the labeled toxin to cells at 4 ◦C. After 1 h, any unbound toxin was removed by washing and the amount of radioactivity associated with the cells was measured. The insets are Scatchard plots of the binding data. These experiments were repeated twice with similar results. (B/F) Bound/Free.

#### 2.1.2. Endocytosis, Recycling, and Degradation

Ebulin l has been shown to be internalized upon binding to glycoproteins and glycolipids containing terminal galactose [16]. To study this process further, COS cells were incubated with 125I-labeled ebulin l or ricin at 37 ◦C for different periods of time. Considering that lactose removes surface-bound toxins but not internalized toxins, toxin internalization was measured as the ratio of endocytosed to surface-bound 125I-labeled

toxin at different time points. As shown in Figure 2a, the amount of internalized ebulin l and ricin levelled off after 30 min, indicating that toxin uptake and intracellular processing approached equilibrium at this time. Approximately 28% of total cell-associated 125I-ebulin l and 23% of total cell-associated 125I-ricin were internalized into the COS cells during a 30 min incubation period at 37 ◦C.

**Figure 2.** (**a**) Kinetics of ebulin l (circles) or ricin (squares) internalization in COS cells: the cells were incubated with 125I-ebulin or 125I-ricin at 37 ◦C for 0 to 120 min and the amount of bound and endocytosed toxins were quantified as described in Section 4.2.2. The data are expressed as the internalized radioactivity in percentage of the total radioactivity associated with the cells. (**b**) Recycling and degradation of ebulin l and ricin in COS cells: the cells were incubated with 125I-ebulin l or 125I-ricin for 20 min at 37 ◦C. Surface-bound toxins were removed by 0.1 M lactose and the incubation continued for the times indicated. Recycling (open circles) was measured as the amount of trichloroacetic acid (TCA)-precipitable toxin in the medium and at the cell surface. Degradation (closed circles) was measured as the amount of radioactivity that could not be precipitated by TCA. Degradation in the presence of bafilomycin A1 (Baf) was measured after 120 min (triangles). In both cases, the data are expressed as a percent of the total radioactivity. The data represent the mean ± SD of two experiments.

We next studied the ability of COS cells to recycle internalized toxins back to the surface and to the culture medium. As shown in Figure 2b, recycling of toxins, measured as trichloroacetic acid (TCA)-insoluble radioactivity in the medium, occurred as a rapid phase that lasted for less than 30 min, followed by a slower phase. Approximately 45% of ebulin l and 70% of ricin were found in the TCA-insoluble fraction after 2 h of incubation at 37 ◦C. We also measured the degradation of toxins internalized by COS cells. As shown in Figure 2b, only 8% of the total internalized 125I-ricin was found to be degraded in COS cells after 2 h incubation at 37 ◦C. However, a higher percentage of ebulin l (36%) was found in the TCA-soluble fraction. Bafilomycin A1, an inhibitor of the vacuolar H+-ATPase [24], inhibited the degradation of ebulin l and ricin, indicating that this process takes place in an acidic compartment. According to this, nigrin b, a nontoxic type 2 RIP, was shown to enter HeLa cells in a similar way to ricin; however, it was much faster and widely degraded. Moreover, the nigrin b-derived material released by cells was completely inactive [21,25].

#### 2.1.3. Mechanism of Endocytosis

Several endocytic mechanisms have been documented, including macropinocytosis, clathrin-dependent endocytosis, caveolae-dependent endocytosis, and clathrin- and caveolae-independent endocytosis [26]. It has been demonstrated that ricin is able to employ different endocytic mechanisms, probably because it can recognize and bind to a great variety of cell surface components [9]. Clathrin-independent endocytosis was first described by studying the uptake of ricin, which continued after the inhibition of clathrin-dependent

endocytosis [27]. To determine if ebulin l and ricin are internalized from clathrin-coated pits in COS cells, the cytosol was acidified to inhibit internalization from clathrin-coated pits [27]. When the cytosolic pH falls below 6.5, clathrin-coated pits at the cell surface can no longer pinch off and form clathrin-coated vesicles. In this study, the cytosol was acidified by preloading the cells with increasing concentrations of NH4Cl followed by its removal [27]. The data in Figure 3a show that, in COS cells, the uptake of ebulin l and ricin were reduced by about 50% after acidification of the cytosol. Control experiments showed that the uptake of transferrin, which is endocytosed by clathrin-dependent endocytosis [28], was reduced by more than 95% under the same conditions. When the cytosol of COS cells was acidified by incubation with acetic acid, similar results to that with NH4Cl pre-pulsing were obtained. Thus, while the endocytosis of transferrin was strongly reduced, there was only an approximately 50% and 55% reduction in the uptake of 125I-ebulin and 125I-ricin, respectively (Figure 3b). This indicates that ebulin l and ricin uptake is mediated by both clathrin-dependent and clathrin-independent endocytosis in COS cells.

**Figure 3.** Effect of acidification of the cytosol on the ability of COS cells to internalize 125I-ebulin l, 125I-ricin, and 125I-transferrin: (**a**) COS cells were incubated for 30 min at 37 ◦C in 4-(2-hydroxyethyl)- 1-piperazineethanesulfonic acid (HEPES) medium pH 7 with the indicated concentrations of NH4Cl. The medium was removed, and a solution containing 0.14 M KCl, 2 mM CaCl2, 1 mM amiloride, 1 mM MgCl2, and 20 mM HEPES, pH 7.0, was added. After 5 min of incubation at 37 ◦C, 125I-ebulin l, 125I-ricin, or 125I-transferrin were added, and cell bound and endocytosed proteins were measured after 20 min of incubation for ebulin l and ricin and after 5 min for transferrin. Symbols: •, ebulin l; , ricin; and , transferrin. (**b**) The cells were incubated for 5 min at 37 ◦C in HEPES medium, pH 5.5, with increasing concentrations of acetic acid. 125I-ebulin l, 125I-ricin, or 125I-transferrin were then added, and after 20 min of incubation for ebulin l and ricin and after 5 min for transferrin, the amount of endocytosed proteins was measured as described above. Symbols: •, ebulin l; , ricin; and , transferrin. The data represent the mean ± SD of two experiments.

#### 2.1.4. Intracellular Transport of Ebulin l in COS Cells

After endocytosis, RIPs move within the endosomal system until they reach the appropriate compartment for entry into the cytosol, where they inactivate ribosomes. To determine the intracellular transport of ebulin l in COS cells, ebulin l was labeled with the fluorophore CY3 and incubated with COS cells. The cellular distribution of ebulin l was studied after incubation at 37 ◦C for various periods of time. The data in Figure 1c demonstrate that, when the cells were treated with ebulin l at 4 ◦C, the toxin is evenly bound all over the cell surface. When the cells were subsequently incubated at 37 ◦C, the amount of ebulin l at the surface was reduced and the fluorescent ebulin l appeared as intracellular dots, suggesting uptake in vesicles (Figure 4). We next performed doublelabeling experiments with EEA1 (early endosome antigen 1), a protein that is associated with early endosomes [29]. As shown in Figure 4A, incubation for 10 min at 37 ◦C resulted in good colocalization of EEA1 and the ebulin l-labeled intracellular structures (Pearson's correlation coefficient (PCC) = 0.42). After 60 min at 37 ◦C, ebulin l remained in vesicles

but did not colocalize to a large extent with EEA1 (data not shown). Instead, some of the structures stained positive for CD63, a late endosome/lysosome marker, colocalized with ebulin l (PCC = 0.24), indicating that the toxin is transported to lysosomes for degradation (Figure 4B). Moreover, ebulin l did not colocalize with mannose-6-phosphate receptor (M6PR), a marker for late endosomes and the trans-Golgi network [30] (PCC = −0.05) (Figure 4C). There is a possibility that only a very small fraction of ebulin l is transported to the Golgi apparatus and that this was not detectable in our assay. However, these results together with further experiments (see below) suggest binding and entry of ebulin l into the endosomal/lysosomal compartment but not to the Golgi apparatus.

**Figure 4.** Transport of fluorescent ebulin l in COS cells and colocalization with markers for different intracellular organelles: the cells were incubated at 4 ◦C for 1 h with CY3-ebulin l to allow binding of the protein and then incubated at 37 ◦C for 10 min (**A**) and 60 min (**B** and **C**) before fixation. Then, the cells were stained with anti-EEA1 (early endosomes) (**A**), anti-CD63 (late endosomes/lysosomes) (**B**), and anti M6PR (Golgi) (**C**) antibodies followed by fluorescein isothiocyanate (FITC)-conjugated secondary antibody. Colocalizations between CY3-ebulin and the different markers were quantified using Pearson's correlation coefficient (PCC). PCC between CY3-ebulin and EEA1 = 0.42 ± 0.08; PCC between CY3-ebulin and CD63 = 0.24 ± 0.07; and PCC between CY3-ebulin and M6PR = −0.05 ± 0.03. The PCC values represent the mean ± SD of 5 images analyzed for each marker. Bar, 50 μm.

Some bacterial toxins, such as diphtheria toxin and anthrax toxin, are translocated to the cytosol from acidic endosomes [31,32]. By contrast, ricin and other bacterial toxins such as cholera toxin and Shiga toxin follow the retrograde pathway from endosomes to the Golgi complex and further to the endoplasmic reticulum before the A chain is translocated to the cytosol [33]. To better understand how ebulin l enters the cytosol, we studied the effects of agents interfering with intracellular routing on the cytotoxic process. Lysosomotrophic amines such as NH4Cl and chloroquine raise the pH within acidic intracellular vesicles. Preincubation of the COS cells with chloroquine and NH4Cl enhanced the cytotoxicity of ebulin l as well as that of ricin (Figure 5). This indicates that ebulin l, similar to ricin, does not require a low pH for translocation to the cytosol. Moreover, the lysosomotrophic amines may stimulate cytotoxicity by preventing toxin degradation by inactivating the lysosomal enzymes, possibly due to an increase in the intralysosomal pH. In addition, we investigated the effect of the fungal inhibitor brefeldin A, which causes Golgi complex disassembly and has been shown to inhibit ricin toxicity. As shown in Figure 5, preincubation of COS cells with brefeldin A markedly reduced as expected the cytotoxicity of ricin but had no significant effect on the cytotoxicity of ebulin l (Figure 5). This indicates that ebulin l follows a Golgi-independent pathway to the cytosol.

**Figure 5.** Effect of brefeldin A, chloroquine, and NH4Cl on protein synthesis in COS cells treated with ebulin l (circles) and ricin (squares): the cells were left untreated (black symbols) or preincubated with 5 μg/mL brefeldin A (crossed symbols), 25 μM cloroquine (grey symbols), and 20 mM NH4Cl (open symbols) for 1 h and then incubated with different concentrations of ebulin l and ricin for 18 h. Protein synthesis was finally measured as indicated in Section 4.2.4. The data represent the mean ± SD of two experiments performed in duplicate.

#### *2.2. Binding, Endocytosis, and Degradation of Ebulin l and Ricin in HeLa Cells Overexpressing dynK44A*

As shown in Figure 3, in COS cells, ebulin l, similar to ricin, was still endocytosed when the formation of clathrin-coated vesicles was inhibited by acidification of the cytosol. This suggests that clathrin-independent endocytosis is responsible for approximately 50% of the ebulin l uptake in those cells. In order to investigate further the mechanism of ebulin l internalization, we used HeLa dynK44A cells, which are HeLa cells with inducible synthesis of a mutant dynamin (K44A) [34]. It has been reported that the GTPase dynamin mediates the scission of clathrin-coated pits and that it is also involved in the budding of caveolae. It has been shown that, in those cells, ricin is internalized by clathrin- and caveolae-independent endocytosis [35]. HeLa dynK44A cells express, under tetracycline regulation, the dominant negative dynamin K44A mutant unable to bind and hydrolyze GTP. When the mutant dynamin is induced by the removal of tetracycline for two days, the cells are defective in clathrin-mediated endocytosis as well as in endocytosis from caveolae [34,36,37]. It has been shown that the prolonged inhibition of clathrin-dependent endocytosis in HeLa dynK44A allows the induction of compensatory mechanisms activating clathrin-independent endocytosis [35,38].

First, we studied the binding of 125I-ebulin l and 125I-ricin and found that HeLa dynK44A cells bound approximately 150 times more ricin than ebulin l. Scatchard analysis indicated that the number of binding sites for ebulin l was approximately 2.5 × 105 and 3.7 × <sup>10</sup><sup>7</sup> for ricin (Figure S1). Cells bound approximately the same amount of 125I-ebulin l independently of mutant expression (data not shown).

We next studied the internalization of ebulin l and ricin both in cells where the mutant dynamin was overexpressed and in cells where its expression was repressed by the presence of tetracycline. Control experiments showed that the endocytosis of 125I-transferrin, which occurs from coated pits [28], was inhibited by more than 90% by overexpression of the mutant dynamin (Figure 6b). By contrast, 125I-ebulin l and 125I-ricin uptake were unchanged in cells expressing the mutant dynamin (Figure 6a). Approximately 25% of total cell-associated 125I-ebulin l and 15% of total cell-associated 125I-ricin were internalized into HeLa dynK44A cells, with and without the induction of mutant dynamin, during 30 min of incubation at 37 ◦C. To study whether the transport of ebulin l to lysosomes was affected by the expression of mutant dynamin, toxin degradation was measured. As shown in Figure 6c, essentially the same degradation rates were obtained whether the

mutant dynamin was expressed by the removal of tetracycline or not. The percentage of ebulin l released by cells in the TCA-soluble fraction was higher (34%) than that of ricin (9%) (Figure 6c). Bafilomycin A1 inhibited toxin degradation, indicating that the process took place in a low-pH compartment (Figure 6c). Moreover, recycling of ebulin l and ricin was not affected by the overexpression of dyn K44A (data not shown) and the values obtained were comparable to those observed in COS cells (Figure 2b). Therefore, the data indicate that the endocytic uptake of ebulin l and ricin, at least in HeLa cells with dynK44A overexpression, does not occur by clathrin-coated pits or caveolae. According to this, it has been shown earlier that ricin endocytosis continued to the same level after dynK44A expression [35]. The cells therefore seem then to be able to upregulate clathrin-independent endocytosis under conditions where a prolonged inhibition of clathrin-dependent endocytosis takes place [35,38].

**Figure 6.** (**a**) Rate of internalization of ebulin l (circles) or ricin (squares) in HeLa dynK44A cells, with (closed symbols) and without (open symbols) the induction of mutant dynamin: HeLa dynK44A cells were grown in the presence or the absence of tetracycline (Tet) for 2 days. The cells were then washed, and 125I-toxins were added. The cells were incubated at 37 ◦C for 0 to 120 min, and bound and endocytosed toxins were quantified as described in Section 4.2.2. (**b**) Endocytosis of 125I-labeled transferrin in HeLa dynK44A cells with (filled bar) and without (open bar) the induction of mutant dynamin: endocytosed transferrin was quantified after 5 min of internalization. (**c**) Degradation of ebulin l and ricin in HeLa dynK44A cells: the cells were grown in the presence (open bar) or the absence (filled bar) of tetracycline for 2 days. The cells were then transferred to a HEPES-containing medium and preincubated without or with (+Baf) bafilomycin A1 (1 mM) for 30 min at 37 ◦C. 125Itoxins were then added, and 20 min later, the surface-bound toxins were removed with a 0.1 M lactose solution at 37 ◦C. The incubation was continued in the presence or in the absence of bafilomycin A1, and after 2 h, further incubation toxin degradation was measured as described in Section 4.2.2. The data represent the mean <sup>±</sup> SD of two experiments. (**d**) Uptake of 125I-transferrin (),125I-ebulin l (•), and 125I-ricin () by HeLa dynK44A cells. The cells grown in the presence of tetracycline were incubated for 5 min at 37 ◦C in HEPES medium, pH 5.5, with increasing concentrations of acetic acid. 125I-transferrin or 125I-toxins were then added as described above, and after 5 and 20 min of incubation, the amount of endocytosed proteins was measured as in (a). The data represent the mean ± SD of two experiments performed in duplicate.

Our results indicated that, after a prolonged inhibition of clathrin-dependent endocytosis in HeLa dyn K44A expressing the mutant dynamin, ebulin l endocytosis continues (Figure 6a), while in COS cells, there was a 50% reduction of the endocytic uptake of ebulin l when endocytosis from coated pits was acutely inhibited by acidification of the cytosol (Figure 3). To test if this is also the case in HeLa dynK44A cells, the uptake of 125I-labeled ebulin l and ricin was measured in cells grown with tetracycline that had been acidified by incubation with acetic acid. Figure 6d shows that, in these cells, the uptake of ebulin l was reduced by about 45% after acidification of the cytosol. The uptake of transferrin was reduced by more than 95% under the same conditions. These results therefore indicate that there is no difference between COS cells and HeLa cells in the way they endocytose ebulin l and ricin. Thus, prolonged inhibition of clathrin-dependent uptake in cells expressing the mutant dynamin can induce an increase in clathrin-independent endocytosis [35,38] while acute inhibition of clathrin-dependent endocytosis by acidifying the cytosol cannot. Endocytosis by mechanisms not involving clathrin-coated pits has been shown for several bacterial toxins such as tetanus toxin, cholera toxin, and plant RIPs such as ricin, lanceolin, and stenodactylin [35,39,40].

#### *2.3. Effect of DynK44A Overexpression and Cytosol Acidification on Ebulin l and Ricin Cytotoxicity*

After endocytosis and transport to the Golgi apparatus and the endoplasmic reticulum, a small number of ricin molecules reach the cytosol, inhibiting protein synthesis. It has been shown earlier that cells overexpressing dynK44A were more resistant to ricin than dynK44A cells expressing endogenous dynamin [35]. The expression of mutant dynamin inhibits transport of endocytosed ricin to the Golgi apparatus, and this transport is important for ricin intoxication [35]. To investigate whether the overexpression of dynK44A changes the ability of ebulin l to intoxicate cells, we measured protein synthesis 18 h after the addition of increasing concentrations of ebulin l or ricin to dynK44A cells grown with and without tetracycline. As shown in Figure 7a, the toxicity of ebulin l was not affected by the expression of mutant dynamin while, as expected, it protected the cells against ricin. Under these conditions, uninduced HeLa dynK44A cells were about 4.8 times more sensitive to ricin that dynK44A-induced cells. In addition, we also studied whether ebulin l and ricin internalized by the clathrin- and caveolae-independent pathway in these cells must be transported through the Golgi apparatus to inhibit protein synthesis. In these experiments, HeLa cells with mutant dynamin were pretreated with brefeldin A, which disrupts the Golgi apparatus, and protein synthesis was measured 18 h later. As shown in Figure 7a and consistent with our previous observations (Figure 5), brefeldin A did not protect cells from the inhibition of protein synthesis by ebulin l but the cells were completely protected against ricin. These data clearly indicate that transport through the Golgi is not required for ebulin l intoxication.

Since the endocytic uptake of ebulin l and ricin in COS cells was reduced by 50% by acidification of the cytosol (Figure 3), we decided to study if toxin internalized under such conditions (by clathrin-independent endocytosis) can intoxicate cells. In these experiments, we used the NH4Cl pre-pulse method to acidify the cytosol. Ebulin l or ricin were added to COS cells, and endocytosis was allowed to proceed for 20 min. Then, the cell surface-bound toxins were removed with a medium containing 0.1 M lactose and the cells were incubated 18 h in normal medium to allow the internalized toxin to intoxicate the cells. After that, the ability of the cells to incorporate [3H] leucine was measured. The data in Figure 7b show that protein synthesis in cells treated in this way was inhibited by ebulin l and ricin to the same extent as in control cells, where endocytic uptake of the RIPs occurred at normal internal pH. The data suggest that, in COS cells, the endocytosis of ebulin l that leads to intoxication of cells takes place predominantly from clathrin-independent mechanisms, and thus, there is no apparent role for clathrin in productive intracellular transport.

**Figure 7.** Ability of ebulin l and ricin to inhibit protein synthesis in HeLa dynK44A (**a**) and COS cells (**b**): dynK44A cells were grown with (closed symbols) and without (open symbols) tetracycline (Tet) for 2 days. Then, the cells were incubated with different concentrations of ebulin l (circles) and ricin (squares) for 18 h, and protein synthesis was measured as indicated in Section 4.2.4. To investigate the effect of brefeldin A (grey symbols) on protein synthesis of K44A cells grown without tetracycline, the cells were preincubated for 1 h with brefeldin A and then incubated with different concentrations of ebulin l and ricin for 18 h, and protein synthesis was measured. The data represent the mean ± SD of two experiments performed in duplicate. (**b**) The toxic effect of ebulin l (circles) or ricin (squares) endocytosed at normal (open symbols) and acidic internal pH (closed symbols): COS cells were incubated for 30 min at 37 ◦C in HEPES medium, pH 7, with and without 25 mM of NH4Cl. The medium was removed, and a solution containing 0.14 M KCl, 2 mM CaCl2, 1 mM amiloride, 1 mM MgCl2, and 20 mM HEPES, pH 7.0, was added. After 5 min of incubation at 37 ◦C, increasing concentrations of ebulin or ricin were added and, after 20 min of further incubation, a growth medium containing 0.1 M lactose was added. The cells were then incubated for 18 h, and protein synthesis was measured. The data represent the mean ± SD of two experiments performed in duplicate.

#### *2.4. Ebulin l Induces Apoptosis in COS Cells*

In addition to rRNA damage, RIPs are capable of inducing cell death by apoptosis [4]. We therefore decided to investigate the death pathways involved in the cytotoxicity of ebulin l in COS cells. Cells treated with ebulin l exhibited the morphological features characteristic of apoptosis such as cell rounding and blebbing (data not shown). To demonstrate the involvement of caspase-dependent apoptosis, caspase-3/7 activation was measured in COS cells exposed to 10−<sup>6</sup> and 10−<sup>7</sup> M ebulin l for 24, 48, and 72 h. As shown in Figure 8a, a time- and dose-dependent activation of effector caspases was observed in COS cells. Caspase activity seems to be significantly induced after treating the cells with 10−<sup>6</sup> M ebulin l for 24 h, and at that concentration, protein synthesis was inhibited by 50% after 18 h (Figure 5). Thus, protein synthesis inhibition seems to be an earlier event than apoptosis in these cells and suggests that apoptosis might be a consequence of the ribotoxic stress induced by the A chain after entry into the cytosol. However, we cannot rule out the possibility that both processes run in parallel. To evaluate the role of the different cytotoxic mechanisms induced by ebulin l, COS cells were pretreated with two inhibitors, the pan-caspase inhibitor Z-VAD, which irreversibly binds to the catalytic site of caspases and was used to selectively inhibit the apoptotic pathway, and the inhibitor of necroptosis, Necrostatin. COS cells were pretreated and maintained in 100 μM Z-VAD, and the cell viability was determined for different ebulin l concentrations. As shown in Figure 8b, caspase inhibition by Z-VAD largely prevented the cytotoxicity of ebulin l after 48 h. At a concentration of 10−<sup>6</sup> M, viability increased from 20% to 72% in the presence of Z-VAD. In contrast, the necroptosis inhibitor Necrostatin did not rescue ebulin l-induced cell death. (Figure 8b). Therefore, these data indicate that apoptosis is the predominant pattern of cell death induced by ebulin l.

**Figure 8.** (**a**) Caspase-3/7 activation in COS cells treated with 10−<sup>6</sup> M (empty bar) or 10−<sup>7</sup> M (filled bar) ebulin l for 24, 48, and 72 h: activity is expressed as the percentage of control values obtained from cells grown in the absence of ebulin l (dashed line). The data represent the mean ± SD of two experiments performed in duplicate. (**b**) Effect of Z-VAD and Necrostatin (Nec-1) on cytotoxicity of ebulin l on COS cells: the cells were preincubated with Z-VAD or Necrostatin for 3 h or left untreated, and then, different concentrations of ebulin l were added and the cells were incubated for 48 h. Cell viability was assessed by a colorimetric assay as indicated in Section 4.2.4. The data represent the mean <sup>±</sup> SD of two experiments performed in triplicate. Symbols: •, untreated; -, +Z-VAD; and , +Nec-1.

#### **3. Conclusions**

This work contributes to elucidating the mechanisms involved in the endocytosis and intracellular transport of the plant toxin ebulin l. Our results demonstrate that ebulin l has a lower number of binding receptors than ricin in COS and HeLa dynK44A cells. This may be due to a change in the structural disposition of the 2γ-subdomain of the ebulin B chain, which limits its ability to bind galactosides. Following binding, ebulin l is internalized by both clathrin-dependent and -independent mechanisms. A short time after internalization, ebulin l is localized to early endosomes and later to lysosomes but apparently not to the Golgi apparatus. The ebulin l molecules that lead to intoxication of cells are internalized via clathrin-independent mechanisms. Moreover, the cytotoxic effect of ebulin l occurs independently of low endosomal pH and does not require transport of the toxin through the Golgi apparatus. Moreover, ebulin l induces a caspase-dependent apoptosis as the predominant cell death mechanism. Importantly, toxins have a potential as therapeutic agents if the toxicity can be targeted to malignant cells. The low unspecific toxicity of ebulin l together with its strong anti-ribosomal activity and induction of apoptosis make it an excellent candidate as a toxic moiety in the construction of immunotoxins and conjugates directed against specific targets. Knowledge of the mechanisms of transport and action of the toxin is essential in achieving this goal.

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

#### *4.1. Reagents and Cells*

The sources of the chemicals and cells were described previously [16,29,34,35,41]. Particular details are given in the Supplementary File S1.

#### *4.2. Methods*

Particular experimental details are given in the Supplementary File S1.

4.2.1. Binding of 125I-Labeled Toxins to Cells and Crosslinking of 125I-ebulin l to Membrane Receptors

Confluent cells were washed twice with ice-cold HEPES medium and incubated at 4 ◦C for 15 min before increasing concentrations of ebulin l or ricin were added. The cells were incubated for 1 h with the toxins and then washed five times with ice-cold phosphate buffered saline (PBS). The radioactivity was measured after dissolving the cells in 0.1 M KOH. Nonspecific binding was estimated by the incubation of cells in the presence of 0.1 M lactose. Receptor dissociation constants and the number of binding sites were estimated by the Scatchard method [42].

Crosslinking of bound 125I-ebulin l to cells was carried out with disuccinimidyl suberate [43].

#### 4.2.2. Measurements of Endocytosis, Recycling, and Degradation

Endocytosis of transferrin was measured after incubation for 5 min with transferrin (100 ng/mL, labeled with 125I) [35]. Internalized toxin (400 ng/mL) was measured as the amount of 125I-labeled toxin that was not removed after incubating the cells with a 0.1 M lactose solution for 5 min at 37 ◦C [44]. Recycling and degradation of toxins were measured as previously described [44].

Endocytosis in cells with acidified cytosol by NH4Cl pre-pulsing or by incubation with acetic acid was assessed as previously described [27]. Cell-bound and endocytosed proteins were measured after 20 min of incubation for 125I-ebulin and 125I-ricin and 5 min for 125I-transferrin.

#### 4.2.3. Immunofluorescence Microscopy

Binding of CY3-ebulin l and double-staining experiments were carried out as previously described [43]. The coverslips were examined with a Zeiss LSM 510 META confocal microscope (Carl Zeiss, Jena, Germany). Colocalization analysis were performed using Coloc2 (version 3.0.0) in Fiji (ImageJ 1.53c) (http://fiji.sc/wiki/index.php/Fiji).

#### 4.2.4. Other Measurements

Protein synthesis was measured as previously described [35]. Cell viability was determined with a colorimetric assay based on cleavage of the tetrazolium salt WST-1 to formazan and the caspase-3/7 activity was assessed by the luminescent assay Caspase-GloTM 3/7 [16].

**Supplementary Materials:** The following are available online at https://www.mdpi.com/2072-665 1/13//102/s1, Figure S1: Toxin binding and Scatchard plots, File S1: Description of the materials and methods.

**Author Contributions:** Conceptualization, L.C., R.I., and J.M.F.; methodology and investigation, R.I., J.M.F., A.L., and L.C.; writing—review and editing, L.C., R.I., J.M.F., and A.L.; funding acquisition, L.C. and A.L. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was funded by grants BIO39/VA39/14 and BIO/VA17/15 (Consejería de Sanidad, Junta de Castilla y León) to L.C. and grant VA033G19 (Consejería de Educación, Junta de Castilla y León) to the GIR ProtIBio. A.L. acknowledges the South-Eastern Norway Regional Health Authority, The Norwegian Cancer Society, and The Norwegian Research Council.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Data are available upon request; please contact the contributing authors.

**Acknowledgments:** We thank Sjur Olsnes for encouragement and unconditional support and Anne Engen for her excellent assistance in cell culture.

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

#### **References**


## *Article* **Sapovaccarin-S1 and -S2, Two Type I RIP Isoforms from the Seeds of** *Saponaria vaccaria* **L.**

**Louisa Schlaak 1, Christoph Weise 2, Benno Kuropka <sup>2</sup> and Alexander Weng 1,\***


**\*** Correspondence: alexander.weng@fu-berlin.de; Tel.: +49-30-838-51265

**Abstract:** Type I ribosome-inactivating proteins (RIPs) are plant toxins that inhibit protein synthesis by exerting rRNA *N*-glycosylase activity (EC 3.2.2.22). Due to the lack of a cell-binding domain, type I RIPs are not target cell-specific. However once linked to antibodies, so called immunotoxins, they are promising candidates for targeted anti-cancer therapy. In this study, sapovaccarin-S1 and -S2, two newly identified type I RIP isoforms differing in only one amino acid, were isolated from the seeds of *Saponaria vaccaria* L. Sapovaccarin-S1 and -S2 were purified using ammonium sulfate precipitation and subsequent cation exchange chromatography. The determined molecular masses of 28,763 Da and 28,793 Da are in the mass range typical for type I RIPs and the identified amino acid sequences are homologous to known type I RIPs such as dianthin 30 and saporin-S6 (79% sequence identity each). Sapovaccarin-S1 and -S2 showed adenine-releasing activity and induced cell death in Huh-7 cells. In comparison to other type I RIPs, sapovaccarin-S1 and -S2 exhibited a higher thermostability as shown by nano-differential scanning calorimetry. These results suggest that sapovaccarin-S1 and -S2 would be optimal candidates for targeted anti-cancer therapy.

**Keywords:** plant toxin; ribosome-inactivating protein (RIP); type I RIP; rRNA glycosylase activity (EC 3.2.2.22); protein isolation; protein sequencing; mass spectrometry

**Key Contribution:** In this work, the complete amino acid sequence of two RIP isoforms from *Saponaria vaccaria* L. was determined and their enzyme properties were characterized.

#### **1. Introduction**

*Saponaria vaccaria* L., also known as cow cockle or prairie carnation, is an annual herbaceous plant belonging to the carnation family (Caryophyllaceae). The flowering plant is a single species in its genus *Vaccaria* and was originally widespread in Eurasia, but nowadays is also found in North and South America, South Africa, and Australia [1]. Many different synonyms for *Saponaria vaccaria* L. exist in the literature, for example *Vaccaria pyramidata* Medik., *Gypsophila vaccaria* (L.) Sm, and *Vaccaria hispanica* (Mill.) Rauschert. For more than 1000 years, Wang-Bu-Liu-Xing (*Vaccariae semen* in Chinese) have been used in traditional Chinese medicine to treat dysmenorrhea, amenorrhea, and lactation failure [1]. The seeds of *Saponaria vaccaria* L. contain triterpenoid saponins including gypsogenin bisdesmosides, cyclic peptides, flavonoids, and polysaccharides [1–4]. In 1995, Bolognesi et al. reported on a 28 kDa type I ribosome-inactivating protein (RIP) from the seeds of *Vaccaria pyramidata* Medik. for the first time [5].

RIPs are distributed in over more than 100 species [6]. However, they have been discovered primarily in the families of Caryophyllaceae, Cucurbitaceae, Euphorbiaceae, Fabaceae, Phytolaccaceae, and Poaceae [6]. RIPs exhibit rRNA *N*-glycosylase activity (EC 3.2.2.22). It was first shown on rat ribosomes that RIPs cleave off a specific adenine residue (A4324) from a conserved GAGA motif in the alpha-sarcin loop of the 28S rRNA [7].

**Citation:** Schlaak, L.; Weise, C.; Kuropka, B.; Weng, A. Sapovaccarin-S1 and -S2, Two Type I RIP Isoforms from the Seeds of *Saponaria vaccaria* L.*Toxins* **2022**, *14*, 449. https://doi.org/10.3390/ toxins14070449

Received: 7 June 2022 Accepted: 27 June 2022 Published: 30 June 2022

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

**Copyright:** © 2022 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 (https:// creativecommons.org/licenses/by/ 4.0/).

As a consequence, the integrity of the 60S ribosomal subunit is compromised, the elongation factor 2 cannot interact with the ribosome any longer and the translation process at the ribosome is stalled [8]. As a result of the cleavage protein synthesis is inhibited, which in the end leads to cell death [8].

A distinction is made between type I and type II RIPs. Type II RIPs such as ricin from *Ricinus communis* L. are composed of an enzymatically active A-chain and a B-chain with lectin-like properties which promotes the binding to galactose molecules on the cell surface [9,10]. That allows them to enter the cell by receptor-mediated endocytosis, using the retrograde transport to the endoplasmic reticulum and reach the cytosol by the endoplasmic-reticulum-associated protein degradation (ERAD) pathway to exert their cytotoxicity [11,12]. Type I RIPs are enzymatically active single-chain proteins [13]. Lacking the B-chain, they enter the cell by receptor-independent endocytosis and accumulate in the late endosomes and lysosomes, where degradation takes place [14]. Only a small quantity escapes from the lysosomes which results in low cytotoxicity [14]. Very prominent and well-characterized type I RIPs are dianthin 30 from *Dianthus caryophyllus* L. and saporin-S6 from *Saponaria officinalis* L. [15,16].

The primary function of RIPs in the plant is not yet completely clear [17]. Different hypotheses on the role of RIPs in plants are under discussion. RIPs possess antiviral, antifungal, antibacterial and insecticidal activities, which may contribute in the protection against plant pests and predators [18,19]. In addition, RIP activity is increased in senescent and stressed leaves, suggesting that RIPs play an important role in regulating the death of plant cells. Also, roles as regulators of protein synthesis and protein storage have been proposed [20].

RIPs have increasingly become a focus of research due to their very efficient protein synthesis inhibitory activity and the resulting high cytotoxicity. Currently, research is being conducted primarily with regard to their promising application in targeted anti-tumor therapy as conjugates with target-specific antibodies, so-called immunotoxins. More than 450 immunotoxins constructed with RIPs are published in the literature [21]. Most of the described immunotoxins have been tested in vitro, some of them also in vivo. Only a few RIP-immunotoxins have been used for clinical studies [21]. Because of the high relevance of this topic, there is a great interest in the scientific community to explore and characterize new RIPs with the aim of identifying additional suitable candidates for immunotoxins. Beyond their potential in anti-cancer therapy, a second area of application for plant extracts containing saponins and RIPs arose in the last few years: the use of RIPs for crop protection. The new application was derived from the function that RIPs have in the plants, in which they are synthesized. Antiviral properties of RIPs have been demonstrated for a wide range of plant viruses [22]. The RIP plant extract can either be directly applied to the plant surface or, alternatively, a virus-resistant transgenic plant may be constructed [23,24]. In terms of crop protection, the antiviral activity of RIPs is complemented by insecticidal effects [19,25].

In 1995, the group of Fiorenzo Stirpe succeeded in isolating a type I RIP from *Saponaria vaccaria* L. (*Vaccaria pyramidata* Medik.) with a size of approximately 28 kDa and an isoelectric point (pI) of 9.5 and could identify its 30 N-terminal amino acids [5]. In 1997, the isolated RIP was used for immunoconjugate construction and successful inhibition of tumor growth in mice by these immunotoxins was demonstrated [26]. The full amino acid sequence, however, has never been published. The authors of the same study reported increased cytotoxicity of the isolated RIP from *Saponaria vaccaria* L. when compared to other RIPs on different cell lines [5]. Due to the increased cytotoxicity, RIPs from *Saponaria vaccaria* L. appear to be promising candidates for use in targeted tumor therapy and crop protection. We now aimed to isolate the type I RIP from *Saponaria vaccaria* L., to determine its precise molecular mass and its complete amino acid sequence, and to characterize its thermal stability and enzymatic activity. Additionally, we attempted to investigate the distribution of type I RIPs across differently processed seed material, as well as identify the main location of type I RIPs within the seed. The main location of type I RIPs within the seed has never been characterized before.

#### **2. Results**

#### *2.1. Protein Extraction from the Seeds of Saponaria vaccaria L.*

Whole dried seeds of *Saponaria vaccaria* L. were available for protein purification. A crude extract with complex protein composition was obtained by aqueous extraction. In SDS-PAGE the crude extract showed promising protein bands in the mass range of 20–35 kDa typical for RIPs (Figure 1A, lane III) [27]. In order to further separate and concentrate these proteins, the aqueous extraction was followed by an ammonium sulfate precipitation at three different concentrations (30%, 60% and 90%). Similar to the crude extract, the three ammonium sulfate fractions also revealed protein bands in the expected mass range for RIPs (Figure 1A, lane IV–VI). The 90%-ammonium sulfate fraction displayed the strongest bands in the mass range of interest (Figure 1A, lane VI). Similar results were obtained in qualitative analysis of the *N*-glycosylase activity. While all three fractions exhibited *N*-glycosylase activity, the 90%-ammonium sulfate fractions possessed by far the highest enzyme activities and therefore were chosen for further protein isolation [28]. These were further subjected to cation exchange chromatography (see Supplementary Figures S1 and S2). At a concentration of 0.3 M NaCl (in 50 mM HEPES, pH 7.0), an apparent single protein eluted with a molecular mass of approximately 25 kDa as estimated from SDS-PAGE (Figure 1A, lane VII). This band had already been among the most intense bands in the crude extract as well as in the 90%-ammonium sulfate fraction. The final yield of protein isolated from the 90% fraction was 8.56 mg per 100 g of whole seeds. We were also able to demonstrate enzyme activity for the isolated protein fraction. High-resolution mass spectrometry of the isolated protein yielded a mono-isotopic mass of 28,763.24 Da and an additional mass peak of 43% relative abundance at 28,793.24 Da (mass difference Δm = 30 Da, Figure 1B), suggesting that the isolated protein fraction contains two RIPs that in the following will be referred to as sapovaccarin-S1 and -S2.

#### *2.2. Protein Sequencing of Sapovaccarin-S1 and -S2*

At the beginning of this study, only the 30 N-terminal amino acids sequenced by Bolognesi et al. had been identified [5]. Here, the determination of the complete amino acid sequence of sapovaccarin-S1 and -S2 was achieved by combining MS-based peptide analysis with PCR experiments. The isolated protein was in-gel digested with trypsin and the peptide map recorded by MALDI-TOF-MS. The resulting peptide mass fingerprint was compared to a protein sequence database using the software Mascot. Since some peptide masses could be matched with known type I RIP sequences, it was confirmed that the isolated protein is a type I RIP. However, many dominant peptide signals from the spectra could not be assigned to any known RIP sequence indicating that the isolated type I RIP has a different sequence. Additionally, the sequences of six tryptic peptides with 64 amino acid residues were determined de novo by MS/MS analysis. Based on the identified peptide sequences, two PCR experiments were performed (see Supplementary Figure S3). DNA of gypsophilin-S, a type I RIP from *Gypsophila elegans* M.Bieb., saporin-S6 and dianthin 30 served as templates for primer design [29–31]. Both experiments combined resulted in a 728-bp DNA sequence that was translated into the 242 amino acid sequence shown in Figure 2.

**Figure 1.** Isolation of sapovaccarin-S1 and -S2 from the seeds of *Saponaria vaccaria* L. and determination of its protein mass. (**A**) SDS-PAGE (12.5%) of the protein purification of sapovaccarin-S1 and -S2 using the whole-seed fraction, Coomassie Brilliant Blue stain. I: Protein marker (in kDa); II: N-terminally His-tagged dianthin (His-dianthin; 0.66 μg), a type I RIP from *Dianthus caryophyllus* L., as a reference; III: Crude extract from *Saponaria vaccaria* L. (4.83 μg); IV: 30%-ammonium sulfate fraction (40.03 μg); V: 60%-ammonium sulfate fraction (45.76 μg); VI: 90%-ammonium sulfate fraction (28.06 μg); VII: The 90%-ammonium sulfate fraction was subjected to cation exchange chromatography. At 0.3 M NaCl sapovaccarin-S1 and -S2 eluted (0.55 μg); VIII: Blank, phosphate buffered saline (PBS). (**B**) High-resolution mass spectrometric analysis of the isolated protein revealed two mono-isotopic masses of 28,763.24 Da and 28,793.24 Da. The y-axis represents the relative signal intensity and the x-axis shows the deconvoluted mass in Da.


**Figure 2.** Protein sequence determination of sapovaccarin-S1 and -S2. Sequence alignment of sapovaccarin-S1 and -S2 and the N-terminus of *Vaccaria pyramidata* Medik. identified by Bolognesi et al. [5]. The final sequences of sapovaccarin-S1 and -S2 are the result of combining MS-based peptide analysis and PCR experiments. Amino acids shown in black were identified by PCR experiments and amino acids (N-terminal and C-terminal region) highlighted in green were not covered by PCR analysis. Peptide sequences identified by MALDI-MS-based peptide sequencing are highlighted by red boxes. Amino acids which could be confirmed by ISD data are underlined in blue. In addition, a triangle is highlighting the amino acid at position 110, indicating the position where sapovaccarin-S1 (alanine) and -S2 (threonine) differ.

This sequence was named sapovaccarin-S1. In the N- and C-terminal region, five and seven amino acids, respectively, could not be identified by the described PCR method (Figure 2). MALDI in-source decay (ISD) was conducted in order to obtain additional information on the N- and C-terminal regions. C-terminal ISD measurements revealed an ion series, enabling to complete the C-terminus (see Supplementary Figure S4). The N-terminus published by Bolognesi et al. matched the sequence identified by PCR and ISD data, finally allowed to confirm the five missing N-terminal residues (Figure 2). The complete protein sequence of mature sapovaccarin-S1 consisting of 254 amino acids is shown in Figure 2.

In addition, in-gel digestions by LysC, chymotrypsin and AspN were performed and peptides were analyzed by LC-ESI-MS. Combining these data sets with the trypsin digest and allowing for unspecific cleavage in the database search, we obtained a sequence coverage of 100% for the sapovaccarin-S1 sequence, thus confirming its correctness. In the trypsin digest, for the peptide with a mass of 2842.4 (pos. 102–126, TVFPEATA**A**NQIVIQYGEDYQSIER) we consistently found a second form with lower signal intensity (roughly one third) with a mass of 2872.4 (Δm = 30 Da) (see Supplementary Figure S5). In the fragment spectrum of this peptide, the b ions beyond b8 and the y ions beyond y16 were shifted by exactly 30 Da, indicative of a substitution A > T at position 110 (see Supplementary Figures S6 and S7). The protein carrying this substitution is referred to as sapovaccarin-S2 (Figure 2). The sapovaccarin-S2 isoform could be detected not only at the peptide level, but also at the DNA level: the codon for A110 revealed two peaks at position 328—a major guanine and a minor adenine peak (see Supplementary Figure S8). The substitution of GCG (sapovaccarin-S1) to ACG (sapovaccarin-S2) implies the amino acid substitution from alanine to threonine at position 110.

As shown in Figure 1B, Orbitrap-based intact protein measurement yielded a monoisotopic mass of 28,763.24 Da, which is almost identical to the theoretical mass calculated from the obtained sapovaccarin-S1 sequence (28,763.12 Da). Additionally, a minor peak was observed (43% relative abundance compared with the main peak) at a monoisotopic mass of 28,793.24 Da which represents sapovaccarin-S2.

The sequence of sapovaccarin-S1 determined here is highly similar to other wellcharacterized type I RIP sequences with 83% sequence identity to gypsophilin-S and 79% sequence identity to each saporin-S6 and dianthin 30, as shown in Figure 3A [29,31,32]. The protein sequence data reported in this paper will appear in the UniProt Knowledgebase under the accession numbers Q7M1L6 and C0HM39.

To predict the protein structure of sapovaccarin-S1 a homology model was built (Figure 3B). The high-resolution crystal structure of dianthin 30 (1.4 Å, PDB ID 1RL0) was chosen as template [33]. The template sequence showed 79% identity and 95% similarity with the target sequence of sapovaccarin-S1—an excellent starting point for homology modelling. The root-mean-square deviation (rmsd) between the template structure and the homology model was 0.52 Å. The low rmsd value indicated a correct global fold of the final model. The protein geometry did not show any Phi-Psi outliers nor atom clashes. The three-dimensional structure of sapovaccarin-S1 corresponded to the common fold of type I RIPs consisting of two domains: the N-terminal domain rich in β-strands and the C-terminal domain rich in α-helices (Figure 3B). The active site was located at the cleft between both domains (Figure 3C). In type I and II RIPs the active-site key residues Tyr73, Tyr121, Glu177 and Arg180 are highly conserved (Figure 3A) [33]. The same applied to sapovaccarin-S1—all key residues were preserved at the same position. The protein surface of sapovaccarin-S1 with its hydrophilic binding pocket is shown in Figure 3C,D. The same structural elements were known from dianthin 30 and saporin-S6 (PDB ID 1QI7) [33,34].


**Figure 3.** Sequence comparison of sapovaccarin-S1 with well-characterized type I RIPs; tertiary structure model of sapovaccarin-S1. (**A**) Sequence alignment of sapovaccarin-S1, saporin-S6, dianthin 30 and gypsophilin-S. Highly conserved active-site resiudes are highlighted in yellow [33]. The alanine at position 110, substitued in sapovaccarin-S2, is highlighted in red. Aligned amino acids labeled with a star symbol (\*) are fully conserved in all four sequences; those with a dot symbol (.) are identical in three out of four sequences and those with colon symbol (:) show moderate identity. Alignment was performed using the Clustal Omega multiple sequence alignment tool (https://www.ebi.ac.uk/Tools/msa/clustalo/, accessed on 11 January 2022). (**B**) Homology model of sapovaccarin-S1 showing its tertiary structure. (**C**) Front view on the protein surface of sapovaccarin-S1. Hydrophilic regions of the protein are highlighted in blue and lipophilic areas in yellow. The substrate binding pocket (red arrow) is located in the cavity in the middle of the protein. (**D**) Back view on the protein surface.The panels (**B**–**D**) have been produced using MOE (version 2020.0901, Chemical Computing Group, Montreal, Canada).

#### *2.3. N-Glycosylase Activity*

Due to their cytotoxic effects type I and type II RIPs have now been studied for over 40 years for their promising use in anti-cancer therapy [35]. The cytotoxic effects of the plant toxins are dependent, on the one hand, on the extent of the endosomal release and on the other hand on their characteristic adenine releasing activity from DNA, RNA, and other polynucleotides [14,36,37]. The group of Fiorenzo Stirpe reported *N*-glycosylase activity on rabbit-reticulocyte lysate and purified rat liver ribosomes for the RIP isolated from *V. pyramidata* [5]. The *N*-glycosylase activity of isolated RIPs from *S. vaccaria* can also be extended to an A30-oligonucleotide substrate (A30). Weng developed an assay, which allows to quantify the enzyme activity of RIPs by measuring the adenine release from an A30-oligonucleotide substrate and the subsequent detection at 260 nm on a thin-layer chromatography (TLC) plate [28]. *N*-glycosylase activity of the isolated sapovaccarin-S1 and -S2 and—for comparative purposes—the isolated gypsophilin-S and recombinant His-dianthin were analyzed with the adenine-releasing assay (Figure 4A). The adenine release of sapovaccarin-S1 and -S2 was higher that of the other investigated RIPs. However, His-dianthin which was recombinantly expressed, released significantly less adenine than the RIPs isolated from the seeds (Figure 4A).

#### *2.4. Cytotoxicity of Sapovaccarin-S1 and -S2*

In order to confirm the previously proven in vitro *N*-glycosylase activity of sapovaccarin-S1 and -S2 in cells, we used label-free live-cell microscopy to investigate the cytotoxicity of sapovaccarin-S1 and -S2 in Huh-7 cells. The confluence analysis of the raw data was performed by the analysis algorithms of the software package CytoSMART. No effect on cell viability could be observed for the different concentrations shown in Figure 4B 24 h after the addition of sapovaccarin-S1 and -S2. After 35 h, 1000 nM sapovaccarin-S1 and -S2 incubation the confluence began to be reduced significantly compared to the control (*t*-test, *p* ≤ 0.05). For 100 nM, 10 nM, and 1 nM the significant reduction in confluence occurred after 46 h, 45 h, and 47 h, respectively (*t*-test, *p* ≤ 0.05). At the end of the sapovaccarin-S1 and -S2 incubation time, a concentration-dependent cytotoxicity of sapovaccarin-S1 and -S2 was observable. A concentration of 1000 nM had the strongest cytotoxic effect compared to the others (*t*-test, *p* ≤ 0.05).

#### *2.5. Thermal Stability*

Protein thermal stability plays a key role in the development of new anti-cancer drugs, for both science and pharmaceutical industrial processes. The thermostability of sapovaccarin-S1 and -S2 was analyzed by nano-differential scanning calorimetry (DSC). In addition, DSC profiles of gypsophilin-S, which has been isolated recently by our group, and recombinantly expressed His-dianthin were recorded as a reference. The DSC profiles and the transition midpoint temperatures (Tm) are presented in Figure 5. Compared to gypsophilin-S (Tm 64.7 ◦C) and His-dianthin (Tm 65.6 ◦C), sapovaccarin-S1 and -S2 had the highest Tm of 68.9 ◦C. Tm values for proteins range typically between 40 and 80 ◦C. Thus, all three RIPs possessed a moderately high thermal stability.

**Figure 4.** Characterization of the enzymatic activity of sapovaccarin-S1 and -S2. (**A**) Quantitative analysis of *N*-glycosylase activity on an A30-oligonucleotide substrate by TLC-densitometry. Sapovaccarin-S1 and -S2 exhibited *N*-glycosylase activity. Gypsophilin-S and His-dianthin were used as positive controls. Sapovaccarin-S1 and -S2 and gypsophilin-S (each 0.01 nM), that were isolated from the seeds, exhibited significantly higher enzymatic activity than His-dianthin (0.01 nM) that was recombinatly expressed. \* significant to sapovaccarin-S1 and -S2 and to gypsophilin-S, t-test, *p* ≤ 0.05. (**B**) Live-cell imaging of Huh-7 cells. After an incubation of 24 h (orange arrow) sapovaccarin-S1 and -S2 was added at different concentrations (0.1–1000 nM). Cells were continuously monitored for 94 h. Sapovaccarin-S1 and -S2 exhibited a concentration-dependent effect on the cell viability of Huh-7 cells.

**Figure 5.** DSC profiles of isolated sapovaccarin-S1 and -S2, isolated gypsophilin-S and recombinantly expressed His-dianthin in PBS, pH 7.4. Each DSC profile was recorded at a protein concentration of 0.4 mg/mL. The transition midpoint temperatures (Tm) were recorded for sapovaccarin-S1 and -S2 at 68.9 ◦C, for gypsophilin-S at 64.7 ◦C and for His-dianthin at 65.6 ◦C. The figure has been produced using NanoAnalyze software (version 3.11.0, TA instruments, New Castle, DE, USA).

#### *2.6. Distribution of Sapovaccarin-S1 and -S2 in Differently Processed Seed Material from Saponaria vaccaria L.*

Beside the whole seeds, the Canadian Carnation Biocompany provided differently processed seed material, which allowed us to study the exact distribution of sapovaccarin-S1 and -S2 in the processed seed material (see Figure 6). The three ammonium sulfate fractions (30%, 60%, and 90% saturation) of each seed fraction were tested for *N*-glycosylase activity. The whole seeds and all eight seed fractions exerted *N*-glycosylase activity for all three ammonium sulfate fractions. With increasing ammonium sulfate concentration, enzyme activity increased—the highest enzyme activities were consistently found in the 90%-ammonium sulfate fractions. The next aim was to quantify the adenine release activity of the differently processed seed material. For this purpose, the 90%-ammonium sulfate fractions of the whole seeds, the fractionated seed material, the embryo-enriched and the perisperm-enriched seed fraction were investigated. The adenine release correlated with the total protein amount. According to the results shown in Table 1, sapovaccarin-S1 and -S2 was most abundant in the perisperm of the seeds of *Saponaria vaccaria* L.

**Figure 6.** Differently processed seed material from *Saponaria vaccaria* L. (**A**) Defatted and grinded seed material. Fractionated seed material was obtained by sieve analysis: the numbers indicate the mesh sizes of the residues of the fractions or if <200 the passage. The embryo-enriched (EEF) and the perisperm-enriched seed fractions (PSF) were obtained by separating the embryo from the rest of the seed using an impact and a roller mill and by sieving. (**B**) Whole seeds in corresponding scale. (**C**) PSF in corresponding scale. (**D**) EEF in corresponding scale.


**Table 1.** Adenine released by the 90%-ammonium sulfate fractions of the differently processed seed material.

#### **3. Discussion**

Here we report the isolation and characterization of sapovaccarin-S1 and -S2, two protein isoforms from *Saponaria vaccaria* L. They were classified as type I RIPs in the course of this study, including their full amino acid sequence. Furthermore, we report for the first time that RIPs in *Saponaria vaccaria* L. are mainly located in the perisperm of the seeds. It should be noted here that the localization of RIPs was determined by evaluating the

*N*-glycosylase activity of extracts from perisperm- and endosperm-enriched seed fractions. The localization in the perisperm might also indicate that sapovaccarin-S1 and -S2 could serve for nitrogen storage in *Saponaria vaccaria* L. This observation might lend some support to the hypothesis that RIPs in addition to having a defense function may also function as storage proteins in some plants. Both RIPs were isolated from the seeds by aqueous extraction, ammonium sulfate precipitation, and cation exchange chromatography. The exact amino acid sequence as well the molecular masses of sapovaccarin-S1 and -S2 were determined by PCR and mass spectrometry. Both isoforms differ only in the substitution of one amino acid at position 110 (Ala110 in sapovaccarin-S1 substituted by Thr110 in sapovaccarin-S2). Their intact masses lie within the characteristic mass range of type I RIPs [27]. The ratio of abundance of sapovaccarin-S1 to sapovaccarin-S2 was consistent in the MS spectrum of intact protein mass, the MS/MS spectrum of tryptic peptides and the DNA chromatogram, indicating that sapovaccarin-S1 is the more abundant isoform.

The occurrence of RIP isoforms within the carnation family was already well described in the literature [15,16,38]. The isoforms described herein differ in only one amino acid: the non-polar alanine in position 110 in sapovaccarin-S1 is substituted by a polar threonine residue in sapovaccarin-S2 provoking a minimal increase in mass (Δm = 30 Da) but no changes in pI (calculated pI 9.87). Hence, separation of both isoforms could be achieved neither by cation exchange chromatography nor SDS-PAGE, resulting in a mixture of the two isoforms. Due to the difference in just one amino acid and the resulting very small mass and non-existent pI differences, the separation of both isoforms by other protein purification methods, such as gel filtration or hydrophobic interaction chromatography, probably could not be achieved. We therefore decided to further investigate the mixture of two isoforms. The decision was reinforced by the fact that the single amino acid substitution was located neither within the active site nor in its immediate vicinity. Although in light of the predicted structures it seems highly unlikely that the substitution A110T will have a major impact on *N*-glycosylase activity, in the future this issue might be clarified by producing the individual isoforms recombinantly by site-directed mutagenesis and determining their activity; however, this would fall outside the scope of the present study.

Type I RIPs are mostly encoded by intron-less genes [39]. Therefore, genomic DNA was used as a PCR template. Even though the N- and C-terminal regions were not determined by PCR sequencing, it can be concluded from the PCR data that sapovaccarin-S1 and -S2 are also encoded by intron-less genes. The isoforms described here could be identified from intact protein mass, tryptic peptides as well as from the chromatogram of the PCR analysis. Given that the amino acid substitution at position 110 can be found in the genomic DNA, the isoforms could not have been created by alternative splicing, but had to have been encoded by two different genes.

RIPs from *Saponaria vaccaria* L. were first mentioned by Bolognesi et al. in 1995 [5]. The authors isolated one RIP by aqueous extraction and cation exchange chromatography. Its molecular mass of 28 kDa was determined by SDS-PAGE and gel filtration. In their study, they determined a pI of >9.5 and demonstrated its *N*-glycosidase activity. These results are essentially in agreement with ours. With the applied isolation and mass determination methods, Bolognesi et al. were not able to differentiate the two isoforms. Their assumption that the isolated protein fraction was composed of one protein, underlines the importance of accurate mass spectrometry methods for future research.

Thermostability of sapovaccarin-S1 and -S2 in comparison to gypsophilin-S and Hisdianthin was studied by DSC. The Tm of sapovaccarin-S1 and -S2, gypsophilin-S and His-dianthin were determined as 69 ◦C, 65 ◦C and 66 ◦C, respectively. Thus, the isolated mixture of sapovaccarin-S1 and -S2 showed the highest thermal stability among proteins studied. Little data on the thermostability of RIPs has been published. Saporin-S6 and gelonin, a type I RIP from *Gelonium multiflorum*, were analyzed by infrared spectroscopy and two-dimensional correlation spectroscopy at neutral pH (50 mM sodium phosphate buffer, pH 7.4) in two different studies [40,41]. Tm of saporin-S6 was measured at 58 ◦C and gelonin's at 66 ◦C, indicating a moderately high thermostability [40,41]. These data are

in accordance with the DSC data of our study. The moderately high thermostability as well as the high *N*-glycosylase activity demonstrated in this study show, that sapovaccarin-S1 and -S2 seem to exhibit potential for toxin moieties in immunotoxins.

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

#### *4.1. Seed Material*

The Canadian Carnation BioProducts Company, Saskatoon, S7H 3R2, Canada provided the seeds of *Saponaria vaccaria* L. (Caryophyllaceae). In addition to whole dried seeds, eight differently processed seed fractions were available: six sieve analysis fractions, an embryo-enriched fraction (EEF) and a perisperm-enriched fraction (PSF). The six sieve analysis fractions were obtained by collecting the particles which were stopped by the sieves (mesh size 40, 60, 80, 100 and 200) and the particles which have passed mesh size 200 (<200). The embryo-enriched fraction (EEF) and the perisperm-enriched fractions (PSF), which is the starch and hull fraction of the seed which remains after the embryo is removed, are achieved by separating the embryo from the rest of the seed using an impact and a roller mill and sieving.

#### *4.2. Isolation of Sapovaccarin-S1 and -S2*

The whole seeds, sieving fraction 40, EEF and PSF had to be ground using an electric mill (M20—Universalmühle, IKA, Staufen, Germany). The raw material of the remaining seed fractions was fine enough from the beginning. Each seed fraction was defatted twice with n-hexane (10 mL/g seed) at 4 ◦C for 30 min and dried at room temperature. The defatted seed fractions were extracted in PBS pH 7.4 (8 mL/g seed) by gentle stirring at 4 ◦C over 24 h and thereafter centrifuged at 8000× *g* for 20 min. The resulting supernatant is referred to as 'crude extract' and was purified by the following steps: Proteins were separated by fractionated ammonium sulfate precipitation at saturations of 30%, 60% and 90% ammonium sulfate. The precipitated proteins were resuspended in PBS and analyzed by SDS-PAGE (12.5%). The protein fractions were also tested for their *N*-glycosylase activity by measuring the released adenine by TLC densitometry at 260 nm [28]. The 90%-ammonium sulfate fractions showed the highest enzyme activity and were used for further purification by cation exchange chromatography using a prepacked SP Sepharose High Performance column (HiTrap SP XL 1 mL, GE Healthcare Europe, Freiburg, Germany) connected to an ÄKTA start protein purification system (GE Healthcare Europe, Freibug, Germany). Then, 7.0 mL of samples equilibrated with 50 mM HEPES, pH 7.0 were applied to the column and bound proteins were eluted from the column by 0.1 M, 0.2 M, 0.3 M, 0.4 M and 1 M NaCl (in 50 mM HEPES, pH 7.0) at a flow rate of 1 mL/min and detected at 280 nm.

#### *4.3. Recombinant Expression of His-Dianthin and Isolation of Gypsophilin-S*

N-terminally His-tagged dianthin 30 (His-dianthin) from the plant *Dianthus caryophyllus* L. was recombinantly expressed in E. coli NiCo21(DE3) (New England Biolabs, Ipswich, QLD, USA), purified by Ni-nitrilotriacetic acid affinity chromatography and analyzed by SDS-PAGE as described elsewhere [42]. Gypsophilin-S was isolated from the seeds of *Gypsophila elegans* M.Bieb. using ammonium sulfate precipitation and subsequent ion exchange chromatography. The isolation is described in detail elsewhere [29].

#### *4.4. SDS-PAGE and Protein Quantification*

12.5% SDS-polyacrylamide gels were used for SDS-PAGE using the Lämmli method. Protein bands were stained with Coomassie Brillant Blue G250 as described elsewhere [43]. Protein concentrations were determined by using a modified Bradford method (ROTI®Nanoquant, Carl Roth GmbH, Karlsruhe, Germany).

#### *4.5. Protein Mass Spectrometry*

Protein and peptide sequences were analyzed by matrix-assisted laser desorption time-of-flight mass spectrometry (MALDI-TOF-MS). All MALDI-TOF-MS measurements were performed with an Ultraflex-II TOF/TOF instrument (Bruker Daltonics, Bremen, Germany) equipped with a 200 Hz solid-state Smart beamTM laser. Data were analyzed using the software FlexAnalysis provided with the instrument. Samples were applied by the dried-droplet technique. Peptides were generated by trypsin, LysC, chymotrypsin and AspN in-gel digestion following a protocol described elsewhere [44]. The mass fingerprints of the generated peptides were recorded in positive reflector mode (RP\_PepMix) over an *m/z* range of 600–4000. α-cyano-4-hydroxycinnamic acid was used as matrix. Selected tryptic peptides got analyzed by tandem MS using the LIFT mode [45]. N-terminal c ions and C-terminal (z + 2) ions were generated from the intact and acetone precipitated protein using in-source decay (ISD). As a matrix, 1,5-diaminonaphthalene (1,5-DAN) was used. Spectra were recorded in positive reflector mode (RP\_PepMix) over an m/z range of 600–6000.

For high resolution intact protein mass analysis by liquid chromatography-electrospray ionization mass spectrometry (LC-ESI-MS) the isolated protein sample was analyzed using the Ultimate 3000 liquid chromatography system connected to a Q Exactive HF mass spectrometer via the ion max source with HESI-II probe (Thermo Scientific, Waltham, MA, USA). The following MS source parameters were used: spray voltage 3.6 kV, capillary temperature 320 ◦C, sheath gas 10, auxiliary gas 4, S-lens RF level 60, intact protein mode on. For the analysis 7 μL of a 10 μM protein solution were desalted and concentrated by injection on a reversed-phase cartridge (MSPac DS-10, 2.1 mm × 10 mm, Thermo Scientific, Waltham, MA, USA) at 60 ◦C using buffer A (0.1% formic acid, 5% acetonitrile in water) at a constant flow rate of 22 μL/min for 3 min. This was followed by a short linear gradient of 5–95% buffer B (0.1% formic acid in 80% acetonitrile, 20% water) within 10 min followed by washing and re-equilibration. Full MS spectra were acquired using the following parameters: mass range m/z 600–2500, resolution 120,000, AGC target 3 × 106, μscans 5, maximum injection time 200 ms. MS raw data were analyzed using BioPharma Finder (version 3.2, Thermo Scientific, Waltham, MA, USA). First, an averaged spectrum over the chromatographic peak was generated followed by spectral deconvolution using a relative abundance threshold of 20% and the function 'consider overlaps' turned off.

#### *4.6. DNA Extraction from the Seeds and Determination of the DNA Sequence by PCR*

Peptide mass fingerprinting and subsequent MS/MS-analysis enabled the identification of first sections of the amino acid sequence of sapovaccarin-S1 and -S2. In order to complete the identified peptides to a full sequence, two PCRs were conducted. Primer pair A was designed based on the peptide mass fingerprint results and primer pair B was derived from the results of the first PCR round and the N-terminus determined by Bolognesi et al. (Table 2). The template DNA was extracted from 75 mg whole dried seeds which were frozen overnight, using the PureLink Plant Total DNA Purification kit (Life Technologies, Carlsbad, CA, USA). DNA concentrations were determined with the NanoDrop (Thermo Fisher Scientific, Waltham, MA, USA). The PCR was conducted using the Phusion High-Fidelity DNA Polymerase (New England Biolabs, Ipswich, QLD, USA). According to the instructions, 25 μL reactions using 64.8 ng of DNA template were prepared. The PCR followed a 3-step protocol with the following cycling steps: Initial denaturation 98 ◦C for 30 s, denaturation 98 ◦C for 30 s, annealing for 30 s (annealing temperature see Table 2), extension 72 ◦C for 22.8 s, final extension 72 ◦C for 10 min. Denaturation, annealing and extension steps were repeated for 35 cycles. PCR products were separated in a 1% agarose gel. The Monarch Genomic DNA purification Kit (New England Biolabs, Ipswich, QLD, USA) was used to extract the PCR products from the gel. For sequencing purposes extracted PCR product concentrations were determined. PCR products were prepared with corresponding primers and sent to LGC Genomics, Berlin, Germany for

sequencing. The Expasy translation tool (https://web.expasy.org/translate/, accessed on 17 September 2021) was used to translate the DNA sequence to a protein sequence.


**Table 2.** Primer pairs used for sequencing of sapovaccarin-S1 and -S2.

#### *4.7. Homology Modeling*

Homology modeling was performed with MOE (version 2020.0901, Chemical Computing Group, Montreal, QC, Canada). The high resolution (1.4 Å) crystal structure of dianthin 30 with PDB ID 1RL0 served as a template). The target sequence and dianthin 30 exhibit 79% sequence identity and 95% sequence similarity. The target sequence and the sequence of dianthin 30 were aligned and checked for correct alignment.

#### *4.8. Adenine-Releasing Assay*

The *N*-glycosylase activity of different samples was determined by using the adeninereleasing assay as described elsewhere [28]. The assay is based on the cleavage of an adenine from an artificial substrate, a DNA oligonucleotide 5- -A30-3- (A30). 10 μL of protein sample was mixed with 10 μg A30 (Metabion International AG, Planegg/Steinkirchen, Germany) and filled up to 50 μL with *N*-glycosylase buffer (50 mM sodium acetate, 100 mM KCl, pH 5). Deviating from the publication the mixtures were incubated at 37 ◦C over night. Samples (each 10 μL) were applied to a TLC 0.25 mm pre-coated silica gel 60 glass plate with fluorescent indicator UV254 (Macherey-Nagel, Düren, Germany) and developed by acetonitrile/water/ammonia (32%), (18:1.6:0.6). In addition, for quantification purposes different adenine amounts (0.125 μg, 0.25 μg, 0.5 μg and 1.0 μg) were applied on the plate. Released adenine was determined by TLC densitometry at 260 nm using the TLCs canner 4 (CAMAG, Berlin, Germany).

#### *4.9. Cytotoxicity*

To monitor the cytotoxicity of sapovaccarin-S1 and -S2 a label-free live-cell imaging system—the CytoSMART Omni system (CytoSMART Technologies B.V., Eindhoven, Netherlands) was used. The CytoSMART Omni system is an automated brightfield microscope, scanning the complete well surface that can be placed in the incubator. The cytotoxicity studies were performed with Huh-7 cells, a hepatocyte carcinoma cell line. The Huh-7 cell line was obtained from Dr. Mirko Pinotti, University of Ferrara. Then, 8000 cells/well were seeded in 96-well plates, each well containing 150 μL Minimum Essential Medium supplemented with 10% FBS, the plate was placed on the CytoSMART Omni device. Cells were cultured at 37 ◦C and 5% carbon dioxide. After 24 h sapovaccarin-S1 and -S2 were added (in 20 μL PBS each 3 wells) ranging from 0.1 to 1000 nM (final concentrations). Control cells were only treated with PBS. Image analysis and confluence calculation was performed using the CytoSMART image analysis software package.

#### *4.10. Differential Scanning Calorimetry*

The thermal stability of protein samples was investigated by differential scanning calorimetry (DSC). A NanoDSC (TA Instruments, New Castle, DE, USA) with capillary cells of 0.3 mL volume was used to carry out the calorimetric measurements. Successive heating and cooling buffer-buffer scans using PBS were repeated three times at a scanning rate of 1 ◦C/min and over a temperature range of 10–100 ◦C. During the measurements a total pressure of 3.0 atm was applied to the reference and the sample cell. Protein samples were prepared in PBS with a concentration of 0.4 mg/mL. Prior to measurements buffers and samples were degassed under vacuum for 15 min. A heating scan of each sample was recorded under the same conditions as the buffer-buffer scans. DSC data was analyzed by using the NanoAnalyze software (version 3.11.0, TA instruments New Castle, DE, USA). Buffer-buffer scans got subtracted from each sample scan.

#### *4.11. Accession Numbers*

The protein sequence data reported in this paper will appear in the UniProt Knowledgebase under the accession numbers Q7M1L6 and C0HM39.

**Supplementary Materials:** The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/toxins14070449/s1. Figure S1: Cation exchange chromatogram of the isolation of sapovaccarin-S1 and -S2 from the 90% ammonium sulfate fraction of perispermenriched seed fraction (PSF). The Y-axis on the left side shows the absorbance at 280 nm in mAU. The Y-axis on the right side represents the composition of the elution buffer in %. 50 mM HEPES, pH 7.0 was used as starting buffer. The NaCl concentration in the elution buffer was gradually increased by adding 1 M NaCl in 50 mM HEPES, pH 7.0. The retention time in min is shown on the X-axis. Sapovaccarin-S1 and -S2 eluted at 0.3 M NaCl. Collected fractions are labeled with 1 to 5; Figure S2: SDS-PAGE (12.5%) of the cation exchange chromatography fractions of sapovaccarin-S1 and S2 using the PSF, Coomassie Brilliant Blue stain. I: Protein marker (in kDa); II: Fraction 1—flow through (6.3–17.5 min); III: Fraction 2 (32.0–35.0 min); IV: Fraction 3 (40.0–42.5 min); V: Fraction 4 (42.5–44.0 min); VI: Fraction 5—sapovaccarin-S1 and -S2 (44.0–46.9 min); VII: Sapovaccarin-S1 and -S2 fraction, concentrated 3 times; VIII: His-dianthin (0.66 μg); Figure S3: DNA sequence of sapovaccarin-S1 obtained by PCR analysis. Based on the tandem MS results, a pair of oligonucleotide primers was designed with the forward primer derived from the DNA sequence of gypsophilin-S (forward primer A) and the reverse primer derived from the C-terminal DNA region of saporin-S6 (reverse primer A). Using these primers and template DNA isolated directly from the seeds of *S. vaccaria* a PCR was performed that yielded a 351-bp PCR product which however did not cover the complete sequence. Therefore, a second primer pair was designed with the reverse primer based on the DNA of the first PCR round (reverse primer B) and the forward primer derived from the DNA of the N-terminus of dianthin 30 (forward primer B). The 449-bp PCR product overlapped in 72 bp with the first PCR product. Combining both sequences resulted in a 728-bp DNA sequence that was translated into a 242-amino acid sequence; Figure S4: In-source decay (ISD) analysis of sapovaccarin (type I RIP from *Saponaria vaccaria* L.). C-terminal ions (z + 2) according to the sequence given in the insert (z9 to z27) and additionally c22 to c28 according to the N-terminal sequence published by Bolognesi et al. are detected; Figure S5: Section of the MS1 spectrum of a tryptic digest of sapovaccarin highlighting the tryptic peptides pos. 102–126 of the isoforms S1 and S2; Figure S6: MS/MS spectra of the tryptic peptide pos. 102–126 of the sapovaccarin isoforms S1 (top) and S2 (bottom). Spectra were analyzed using the Mascot MS/MS search software and the corresponding peptide sequences and the matched b- and y-ions are indicated; Figure S7: Theoretical b- and y-ion series of the tryptic peptide pos. 102–126 of the sapovaccarin isoforms S1 (left) and S2 (right). B-ions differ by 30 Da starting from b9 and y-ions starting from y17; Figure S8: Sequence chromatogram of nucleotides encoding for amino acid positions 107 to 113. At nucleotide position 328 the guanine peak overlays an adenine peak. The codon change from GCG (sapovaccarin-S1) to ACG (sapovaccarin-S2) implicates an amino acid substitution at position 110 from alanine to threonine. Consistent with the results of the trypsin digest and the Orbitrap-based intact protein measurement, the guanine peak is more intense than the adenine peak.

**Author Contributions:** L.S. designed and performed experiments, analyzed data and wrote the main manuscript. C.W. and B.K. designed and performed mass spectrometry experiments, analyzed data, wrote the mass spectrometry parts of the manuscript, read and edited the manuscript; A.W. designed research, performed cytotoxicity experiments, analyzed data, read and edited the manuscript. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation)—project number 422686308. We acknowledge support by the Open Access Publication Fund of the Freie Universität Berlin.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Protein sequencing data have been deposited in UniProt Knowledgebase under the accession numbers Q7M1L6 and C0HM39.

**Acknowledgments:** For mass spectrometry (C.W. and B.K.), we would like to acknowledge the assistance of the Core Facility BioSupraMol supported by the Deutsche Forschungsgemeinschaft (DFG). For the provided seed material, we would like to acknowledge Canadian Carnation BioProducts Company.

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

#### **References**


## *Article* **Isolation, Characterization and Biological Action of Type-1 Ribosome-Inactivating Proteins from Tissues of** *Salsola soda* **L.**

**Nicola Landi 1,†, Sara Ragucci 1,†, Lucía Citores 2, Angela Clemente 1, Hafiza Z. F. Hussain 1, Rosario Iglesias 2, José M. Ferreras <sup>2</sup> and Antimo Di Maro 1,\***


**Abstract:** Ribosome-inactivating proteins (RIPs) are known as RNA N-glycosylases. They depurinate the major rRNA, damaging ribosomes and inhibiting protein synthesis. Here, new single-chain (type-1) RIPs named sodins were isolated from the seeds (five proteins), edible leaves (one protein) and roots (one protein) of *Salsola soda* L. Sodins are able to release Endo's fragment when incubated with rabbit and yeast ribosomes and inhibit protein synthesis in cell-free systems (IC50 = 4.83–79.31 pM). In addition, sodin 5, the major form isolated from seeds, as well as sodin eL and sodin R, isolated from edible leaves and roots, respectively, display polynucleotide:adenosine glycosylase activity and are cytotoxic towards the Hela and COLO 320 cell lines (IC50 = 0.41–1200 nM), inducing apoptosis. The further characterization of sodin 5 reveals that this enzyme shows a secondary structure similar to other type-1 RIPs and a higher melting temperature (Tm = 76.03 ± 0.30 ◦C) and is nonglycosylated, as other sodins are. Finally, we proved that sodin 5 possesses antifungal activity against *Penicillium digitatum*.

**Keywords:** antifungal activity; agretti; cytotoxicity; edible plants; protein purification; rRNA N-glycosylases

**Key Contribution:** Here, we reported the isolation and characterization of seven type-1 RIPs named sodins from *Salsola soda*, known for its edible leaves (agretti in Italian). Furthermore, we focused our attention on their biological and antifungal activities.

#### **1. Introduction**

Ribosome-inactivating proteins (RIPs) are a group of ribotoxic enzymes which act on ribosomes in a highly specific and irreversible manner. They are N-glycosylases (EC 3.2.2.22) capable of hydrolyzing the N-glycosidic bond of a specific adenosine in the sarcin ricin loop (SRL) of major rRNA (A4324, rat liver 28S rRNA numbering) [1]. The consequent formation of an apurinic site prevents the interaction of prokaryotic or eukaryotic elongation factors (EF-G or EF-2, respectively) with ribosomes, blocking mRNA-tRNA translocation and thus inhibiting protein synthesis and inducing the apoptotic pathway [2]. In addition, these enzymes possess polynucleotide:adenosine glycosylase (PNAG) activity on different polynucleotide substrates (e.g., viral RNA and herring sperm DNA) [3–5]. On the other hand, other enzymatic activities such as DNAse [6,7], RNAse [8], chitinase [9], phosphatase, lipase [10] and superoxide dismutase properties [11–13] have also been attributed to RIPs, although some authors ascribe these activities to possible contamination [4,14]. RIPs are mostly found in flowering plants [15,16], few are found in fungi [17] and bacteria [18] and one is found in alga [19].

**Citation:** Landi, N.; Ragucci, S.; Citores, L.; Clemente, A.; Hussain, H.Z.F.; Iglesias, R.; Ferreras, J.M.; Di Maro, A. Isolation, Characterization and Biological Action of Type-1 Ribosome-Inactivating Proteins from Tissues of *Salsola soda* L.. *Toxins* **2022**, *14*, 566. https://doi.org/10.3390/ toxins14080566

Received: 22 July 2022 Accepted: 17 August 2022 Published: 19 August 2022

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

**Copyright:** © 2022 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 (https:// creativecommons.org/licenses/by/ 4.0/).

Based on the presence or absence of a quaternary structure, there are two main groups of RIPs: single-chain proteins (type-1 RIPs), with a molecular weight of ~30 kDa and basic pI, and two-chain proteins (type-2 RIPs), with a molecular weight of ~60 kDa and neutral pI. The latter consist of an enzymatic active A-chain linked through a disulphide bridge to a lectinic B-chain, which allows for the entry into the cell. For this reason, type-2 RIPs are more toxic with respect to type-1 RIPs. In addition, a third group of non-canonical single-chain RIPs (type-3 RIPs) was found only in Poaceae, including JIP60 isolated from barley [20,21] and b-32 isolated from maize [22], made of a type-1-like N-terminal domain with N-glycosylase activity, covalently linked to a C-terminal domain with an unknown function [23].

Although their physiological role is still unknown, RIPs have a broad spectrum of activities, including antiviral, antibacterial and antifungal action, as well as anticancer properties [24–26]. Thus, the potential applications of RIPs span many fields, from agriculture (e.g., toxicity to pests and antifungal activity) [27] to biomedicine for the construction of antibody-RIPs conjugates (i.e., immunotoxins) against target cancer cells [28]. From the foregoing, it is clear that continuing the study of RIPs distribution in higher plants (including edible species) can contribute to expanding the availability on novel potential biotechnological and pharmacological tools.

*Salsola soda* L., commonly known as barilla plant or 'agretti' in Italy, is an annual, edible halophytic plant that is widespread in south Europe, mostly near the coast. In the past, the plant was used for the production of an impure sodium carbonate named 'barilla' from the sodium chloride in the soil (e.g., to make soap and glass) [29]. The plant tissues are rich in alkaloids, saponins, coumarins and sterols, as well as flavonoids with anti-inflammatory and antidiabetic potential [30]. According to the taxonomy, *S. soda* belongs to Caryophyllales [31], a plant order known as a source of RIPs [32]. The plant is an Amaranthacea with succulent leaves, small sessile hermaphrodite flowers and indehiscent fruits. It is known that the isoforms of these ribonucleolytic enzymes can be found in different tissues of the same plant [33–35] or in the same plant tissue [36]. However, little information about the distribution of RIPs in plant organs and tissues is available in the literature, especially with regard to edible plants, considering that they are often consumed raw [37].

In this framework, we report the purification and characterization of seven novel RIPs named sodins: five from the seeds, one from the roots and one from the edible leaves of *S. soda*. We describe the distribution of both N-glycosylase and PNAG activity in the different tissues of *S. soda*. Moreover, considering the high expression of sodins in the seeds, we further characterized the major form (i.e., sodin 5) by evaluating some structural features, the cytotoxic effect in cancer cell lines and the antifungal activity against *Penicillium digitatum*.

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

#### *2.1. Purification of Type-1 RIPs from Salsola soda Seeds*

The crude extract from *S. soda* seeds, obtained by homogenizing seeds (50 g) in 500 mL of phosphate-buffered saline, pH 7.2, possessed the ability to depurinate the hsDNA substrate (data not shown) [38]. Thus, to ascertain the presence of type-1 RIPs, the total crude extract was subjected to acid precipitation following a protocol for the extraction of basic proteins [34]. The supernatant was subjected to step-wise cation exchange chromatography, and eluted basic proteins were separated by gel-filtration. Pooled fractions with a molecular weight around 29 kDa were further subjected to cation exchange chromatography on the CM-Sepharose column, eluted with a linear NaCl gradient (0–0.17 M) in 5 mM Na-phosphate buffer. As shown in Figure 1A, five protein peaks with PNAG activity were detected, confirming the presence of various PNAG-enzymes (isoforms), which is a common feature in plant RIP expression [15].

**Figure 1.** (**A**) Elution profile after cation exchange chromatography on the CM-Sepharose column, showing five peaks (peaks 1–4 and sodin 5) with PNAG activity (arbitrary units). (**B**) SDS-PAGE analysis of 194–201 fractions (5.0 μg) from sodin 5 obtained after cation exchange chromatography (**A**). M, molecular weight markers. SDS-PAGE in the presence of β-mercaptoethanol was carried out in 12% polyacrylamide separating gel and then stained with Coomassie brilliant blue.

In this framework, we decided to first characterize the main protein peak, eluted at a higher ionic strength (peak 5). In particular, the fractions from 194 to 201, corresponding to the main peak (hereafter, sodin 5), were evaluated by SDS-PAGE, showing a single protein band with an electrophoretic migration of ~29 kDa (Figure 1B). Thus, the fractions were pooled, dialyzed and used for further enzymatic and structural characterization.

The purification yield obtained by this procedure was of about 2.9 ± 0.15 mg/100 g of seeds for sodin 5.

#### *2.2. Enzymatic and Structural Features of Sodin 5*

In order to ascertain that the in vitro protein synthesis inhibition and the PNAG activity of sodin 5 were due to N-β-glycosylase action (characteristic enzymatic hallmark of RIPs from plants), we tested the β-fragment release by incubating the protein with rabbit ribosomes (Endo's assay) [32]. As shown in Figure 2A, sodin 5 is able to deadenylate rRNA from rabbit reticulocyte lysate, releasing the β-fragment after aniline treatment. Furthermore, as demonstrated by the PNAG activity assay, sodin 5 is ~1.8-fold less active than quinoin, a type-1 RIP isolated from *Chenopodium quinoa* seeds and used as a reference PNAG-enzyme (Figure 2B) [39].

In addition, sodin 5 purified from *S. soda* seeds inhibited protein synthesis in a rabbit reticulocyte lysate system, with an IC50 value of 4.83 pM (0.14 ng/mL). This value is similar to that of quinoin (IC50= 5.08 pM; 0.15 ng/mL) and ~7.7-fold lower than that of saporin S6 (IC50 = 37 pM; 1.09 ng/mL) type-1 RIP isolated from *Saponaria officinalis* seeds [40]. On the other hand, the IC50 value of sodin 5 is of keen interest considering that type-1 RIPs have values of IC50 between 10 and 4000 pM [41].

The high amount of sodin 5 obtained from seeds of *S. soda* allowed us to perform a study on its secondary structure by Circular Dichroism (CD-) analysis. The far UV CD-spectrum of sodin 5 suggested that the periodic secondary structure of the protein is partially dominated by the α-helix, with a predicted percentage of ~30% (~25% β-strand) (Figure 3A). Therefore, these data show that sodin 5 has a secondary structure content similar to other RIPs, sharing a common 3D fold (RIP fold) consisting of a β-sheet Nterminal domain and an α-helix-rich C-terminal domain [42,43]. Subsequently, the thermal denaturation curve of sodin 5 was obtained using UV-spectroscopy by measuring the increment of absorbance at 278 nm, increasing the temperature. The melting temperature

(Tm) of sodin 5 was 76.03 ± 0.30 ◦C (Figure 3B). The thermal unfolding curve at pH 7.2 shows that this type-1 RIP is a highly stable protein. In particular, the Tm value of sodin 5 is higher than that of both quinoin (68.2 ◦C [39]) and saporin S6 (58.0 ◦C) [44]. Both quinoin and saporin S6 have been isolated from Caryophyllales, the same plant order of *S. soda*.

**Figure 2.** (**A**) rRNA N-glycosylase activity on rabbit ribosomes. Quinoin from *C. quinoa* seeds (3.0 μg; lanes 3 and 4) as a positive control and sodin 5 (3.0 μg; lanes 5 and 6) were incubated with ribosomes. Then, rRNA was extracted, treated with acid aniline and separated as described in the Materials and Methods section. (+) and (−) indicate with and without aniline treatment. 'β-frag' indicates the position of Endo's fragment released by the aniline treatment of rRNA from rabbit ribosomes. (**B**) Polynucleotide:adenosine glycosylase activity of BSA (negative control) or quinoin and sodin 5 type-1 RIPs. Proteins (3.0 μg) were assayed on salmon sperm DNA as described in the Materials and Methods section. The mean results ± SD of three experiments performed in triplicate are reported. The data were compared to the control and analyzed by one-way ANOVA with Dunnett's post hoc test (\*\*\*\*, *p* < 0.0001).

**Figure 3.** (**A**) Far-UV CD spectrum of sodin 5. (**B**) Thermal denaturation profile of sodin 5 (concentration: 0.15 mg mL−1). The fraction unfolded at 278 nm is plotted as a function of temperature. The red line represents fit curve.

Finally, considering that several type-1 RIPs are N-glycosylated, a specific analysis for glycoproteins detection was carried out. When analyzed by SDS-PAGE and sugar staining, sodin 5 appeared to be non-glycosylated (Figure S1).

#### *2.3. Minor Forms of Type-1 RIPs from Salsola soda Seeds*

To obtain information on minor peaks 1–4 (Figure 1A), showing PNAG activity and eluted at a lower ionic strength with respect to sodin 5, the fractions of each protein peak were analytically re-chromatographed by FPLC on an AKTA Purifier System from cation exchange chromatography using a Source 15S PE 4.6/100 column (Figure S2A). Each eluted peak displayed a single band of ~29 kDa by SDS-PAGE analysis (Figure S1A).

The pooled fractions of peaks 1–4, (hereafter sodins 1–4) were able to release the β-fragment similarly to sodin 5 as a consequence of the RIPs action (Figure S3). Therefore, sodins 1–4, with a molecular weight of ~29 kDa and specific N-β-glycosylase activity, can be considered type-1 RIPs. Moreover, among type-1 RIPs from *S. soda* seeds, sodin 1 displayed the higher PNAG activity (~2.3-fold more active than sodin 5), while sodins 2, 3 and 4 are, respectively, ~1.3-, 2.0- and 1.8-fold less active than sodin 5 (Figure S2B). These data agreed with the documented different ability of type-1 RIPs to deadenylate nucleic acid substrates [3].

In addition, the specific analysis for glycoproteins detection after SDS-PAGE was carried out. Similar to sodin 5, the analysis shows that these enzymes are non-glycosylated (Figure S1B).

The purification yield obtained by this procedure was of about 0.37 ± 0.01, 0.45 ± 0.02, 0.47 ± 0.01 and 0.67 ± 0.02 mg/100 g of seeds for sodins 1–4, respectively.

#### *2.4. Type-1 RIPs from Edible Leaves and Roots of Salsola soda*

In order to evaluate the RIPs distribution in the tissues of *S. soda*, the same protocol used for the purification of sodins from *S. soda* seeds was also applied on the edible leaves and roots of this plant. The approach, coupled with the detection of enzymatic activity, allowed for the purification of two different type-1 RIPs from *S. soda* edible leaves and roots, respectively. However, considering the lower number of raw basic proteins, after cation step-wise chromatography and gel-filtration (see Section 4.2), the pools of basic proteins from the edible leaves and roots with a molecular weight of 29 kDa were separately subjected to analytical cation exchange chromatography using a Source 15S PE 4.6/100 column on AKTA Purifier System (Figure 4A).

The single protein peaks from edible leaves and roots were analyzed by SDS-PAGE to verify the purity and integrity. As reported in Figure 4B, both protein peaks showed the presence of a single band with an electrophoretic migration of ~29 kDa. These two-novel type-1 RIPs from edible leaves and roots were named sodin eL and sodin R, respectively.

The purification yield obtained by this procedure was of about 17.5 ± 0.61 μg/100 g of edible leaves and 27.8 ± 0.87 μg/100 g of roots for sodin eL and sodin R, respectively, confirming the low number of type-1 RIPs in edible leaves and roots compared to the quantity found in seeds.

In addition, the specific analysis for glycoproteins detection after SDS-PAGE shows that these enzymes are non-glycosylated, like both sodin 5 and sodins 1–4 from *S. soda* seeds (Figure S4).

Finally, the N-β-glycosylase action (characteristic enzymatic hallmark of RIPs from plants) of sodin eL and sodin R has been tested. As shown in Figure 5A, both proteins release the β-fragment by incubating the protein with rabbit ribosomes (Endo's assay) following aniline treatment, similarly to sodin 5 from *S. soda* seeds. In addition, type-1 RIPs from *S. soda* roots and edible leaves also displayed PNAG activity. In particular, as shown in Figure 5B, sodin eL and sodin R displayed a PNAG activity that was ~2.2- and 2.9-fold higher than that of sodin 5.

**Figure 4.** (**A**) Elution profile after FPLC on an AKTA Purifier System from cation exchange chromatography using a Source 15S PE 4.6/100 column, showing a single protein peak from *S. soda* roots and edible leaves, named sodin R and sodin eL, respectively. The elution profile of sodin 5 from *S. soda* seeds was reported as a reference chromatographic profile. (**B**) SDS-PAGE analysis of fractions (5.0 μg) from sodin 5, sodin R and sodin eL (lanes 1, 2 and 3, respectively), obtained after Source 15S chromatography. M, molecular weight markers. SDS-PAGE in the presence of β-mercaptoethanol was carried out in 12% polyacrylamide separating gel and then stained with Coomassie brilliant blue.

**Figure 5.** (**A**) rRNA N-glycosylase activity assayed on rabbit ribosomes. Sodin 5 (3.0 μg; lanes 3 and 4) as a positive control and sodin eL (3.0 μg; lanes 5 and 6) or sodin R (3.0 μg; lanes 7 and 8) were incubated with ribosomes. Then, rRNA was extracted, treated with acid aniline and separated as

described in the Materials and Methods section. (+) and (−) indicate with and without aniline treatment. 'β-frag' indicates the position of Endo's fragment released by the aniline treatment of rRNA from rabbit ribosomes. (**B**) Polynucleotide:adenosine glycosylase activity of BSA (negative control) or sodin 5, sodin eL and sodin R type-1 RIPs. Proteins (3.0 μg) were assayed on salmon sperm DNA as described in the Materials and Methods section. The mean results ± SD of three experiments performed in triplicate are reported. Data were compared to the control and analyzed by one-way ANOVA with Dunnett's post hoc test (\*\*\*\*, *p* < 0.0001).

In addition, since the rRNA N-glycosylase activity might play a role in plant defense, we assayed the effect of sodin 5, sodin eL and sodin R, as well as quinoin, on ribosomes from yeasts (*Saccharomyces cerevisiae*) homologous to ribosomes from the putative fungal pathogens of plants. As shown in Figure 6, these RIPs displayed rRNA N-glycosylase activity on yeast ribosomes, as indicated by the release of a diagnostic β-fragment identical to that reported for BE27 from *Beta vulgaris* L. (sugar beet) and type-1 RIPs from *Phytolacca dioica* L. [45,46].

**Figure 6.** rRNA N-glycosylase activity assayed on yeast ribosomes. rRNA N-glycosylase activity was analyzed as reported in the Materials and Methods section. Each lane contained 5 μg of RNA isolated from either untreated (control) or RIP-treated ribosomes from yeast. (+) and (−) indicate with and without aniline treatment. 'β-frag' indicates the position of Endo's fragment released by the aniline treatment of rRNA from yeast ribosomes.

Finally, sodin eL and sodin R inhibited protein synthesis in a rabbit reticulocyte lysate system, with IC50 values of 79.31 pM (2.3 ng/mL) and 65.52 pM (1.9 ng/mL), respectively. These values are ~15-fold higher with respect to the IC50 of sodin 5 isolated from *S. soda* seeds.

#### *2.5. Cytotoxic Effects of Sodins from S. soda Tissues in Cell Cultures*

RIPs are cytotoxic toward several human cell lines (normal and malignant), although type-1 RIPs are usually less cytotoxic than type-2 RIPs due to the lack of a B-chain, which improves the entry of the A-chain in the cells. Indeed, typical IC50 values of toxic type-2 RIPs on cultured animal cells are in the range of 0.3–17,000 pM, while IC50 values of type-1 RIPs are in the range of 170–3300 nM [41]. Thus, we decided to verify the cytotoxic effects of sodins or quinoin on both HeLa and COLO 320 cell lines.

Table 1 lists the IC50 values (concentration of protein causing the death of 50% of cells) of sodin 5, sodin eL and sodin R from *S. soda* seeds, edible leaves and roots, respectively, compared with the IC50 of quinoin. Type-1 RIPs from both *S. soda* tissues and *C. quinoa*

seeds were toxic to HeLa and COLO 320 cells, exhibiting IC50 values ranging from 0.41 to 1200 nM.

**Table 1.** Cytotoxicity of sodins and quinoin. HeLa or COLO 320 cells were grown in RPMI 1640 medium and incubated with different RIP concentrations for 48 or 72 h, and cell viability was evaluated by a colorimetric assay, as indicated in Section 4.6 of the Materials and Methods section. When reported, cells were pre-treated with Z-VAD for 3 h (see Materials and Methods) and then incubated with different RIP concentrations. Data represent the mean of three experiments performed in triplicate.


The most sensitive were HeLa cells, with IC50 values from 0.41 to 2.0 nM after 48 h of treatment, while COLO 320 cells have values between 160 - >120 nM after 72 h of treatment (Figure 7A). These data agree with those previously reported for other type-1 RIPs, such as type-1 RIPs isolated from *P. dioica*, which exhibit IC50 values ranging from 1.0 to 1000 nM against the same cell lines [45]. On the other hand, the cytotoxicity of sodin 5 and quinoin is similar, with IC50 values 103-fold higher for COLO 320 cells with respect to Hela cells.

There are important differences in toxicity among type-1 RIPs based on their capability to reach the ribosomes of target cells. Based on the above studies, sodins and quinoin display a considerable toxicity against HeLa cells for type-1 RIPs. To see if sodin 5 and sodin R can reach and inactivate the ribosomes after being endocytosed, we analyzed the rRNA from HeLa cells after 48 h of RIP treatment. Figure 7B displayed that the ribosomes were depurinated, as proved by the detection of a diagnostic β-fragment following RNA treatment with acid aniline. Thus, both sodin 5 and sodin R can reach the ribosomes of target cells, inhibiting protein synthesis.

Several studies highlight that RIP cytotoxicity in the cells is associated with their ability to induce apoptosis [12]. Apoptosis might be a consequence of the ribotoxic stress induced by the RIP after entry into the cytosol, or both processes could run in parallel. Apoptosis is characterized by cell shrinkage, nuclear condensation, changes in the cell membrane and mitochondria, DNA fragmentation into 200 base oligomers and protein degradation by caspases. In this framework, in order to ascertain if the observed cytotoxic effects of both sodin 5 and quinoin were mediated via apoptosis, we evaluated the sensitivity to the pan-caspase inhibitor Z-VAD or the cleavage of chromosomal DNA into oligonucleosomal fragments, considered a late-stage apoptosis hallmark. In particular, HeLa cells were pretreated and maintained in 100 μM Z-VAD for 48 h, and the cell viability was determined for different sodin 5 and quinoin concentrations. As shown in Figure 7A, the presence of Z-VAD improved cell survival. In particular, in the presence of Z-VAD, viability increased from 14 to 49% in 4.3 × <sup>10</sup>−<sup>7</sup> M sodin 5-treated cells and from 14 to 46% in 5.7 × <sup>10</sup>−<sup>7</sup> M quinointreated cells. On the other hand, when COLO 320 cells were treated for 72 h with RIP concentrations close to their IC50, the breakdown of the nuclear DNA into oligonucleosomal fragments was clearly observed (Figure 7C).

**Figure 7.** Induction of cytotoxicity and apoptosis on HeLa and COLO 320 cells by sodins and quinoin (**A**). Effect of sodins or quinoin on the viability of HeLa (left panel) and COLO 320 (right panel) cells. Cells were grown in RPMI 1640 medium and incubated with different type-1 RIP concentrations for 48 h (HeLa) and 72 h (COLO 320), and cell viability was evaluated by a colorimetric assay, as indicated in Section 4.6 of the Materials and Methods section. To investigate the effect of Z-VAD on the viability of HeLa cells, the cells were preincubated for 3 h with Z-VAD and then incubated with different concentrations of sodin 5 or quinoin for 48 h, and cell viability was evaluated. Data represent the mean ± SD of two experiments performed in duplicate. (**B**) rRNA N-glycosylase activity of sodin 5 and sodin R on RNA from HeLa cells. rRNA N-glycosylase activity was evaluated as reported in the Materials and Methods section. Each lane contained 2.0 μg of RNA isolated from either untreated cells (C, control) or cells incubated with 8 nM of sodin 5 or 5 nM of sodin R for 48 h. The arrow indicates the RNA fragment released as a result of RIP action upon the acid aniline treatment. Numbers indicate the size of the standards (M) in nucleotides. (**C**) Effect of sodin 5 and quinoin on internucleosomal DNA fragmentation. COLO 320 cells were incubated in the absence (C, control) or presence of 0.4 μM of sodin 5 or 0.6 μM of quinoin (Q) for 72 h. The DNA was isolated, and 4.0 μg was electrophoresed, as indicated in Section 4.7. The numbers indicate the corresponding size of the standards (M) (λDNA HindIII/EcoRI) in Kb. (+) and (−) indicate with and without aniline treatment.

Overall, our data suggested that the apoptotic pathway was implicated in the cell death mediated by sodin 5 and quinoin, as already proved for other type-1 RIPs [41].

#### *2.6. Effect of Sodin 5 and Quinoin on the Growth of P. digitatum*

A potential role for RIPs as plant defense proteins has been proposed based on their enzymatic activity, which can act by either inactivating pathogen ribosomes or their own ribosomes, causing cell death [27]. Antifungal activity has been attributed to several RIPs. In particular, a strong antifungal activity against *P. digitatum* has been described for the apoplastic type-1 RIP beetin 27 (BE27) from sugar beet and for the type-1 RIPs PD-S2 and dioicin 2 from *P. dioica* [45,47]. *P. digitatum* is a necrotrophic fungus responsible for the postharvest decay of citrus, an economically important crop worldwide. Therefore, we carried out experiments to evaluate the effects of sodin 5 from *S. soda* seeds and quinoin from *C. quinoa* seeds on the growth of *P. digitatum*. Thus, conidia of *P. digitatum* were grown in PDB medium for 24 h before exposure to different RIP concentrations, continuing the treatment for a further 46 h. As shown in Figure 8, sodin 5 and quinoin reduced the fungal growth in a concentration-dependent manner. Both RIPs induced a strong decrease in the growth at 40 μg/mL. Thus, 40, 10 and 4.0 μg/mL of sodin 5 resulted in 70%, 34% and 6% growth inhibition, respectively, after 70 h of growth. Similar results were obtained for quinoin, with 61%, 27% and 10% growth inhibition at the same concentrations. Sodin 5 and quinoin added from the beginning to the conidia as starting material inhibited fungal growth to the same extent as RIPs added at 24 h (once conidial germination occurred; data not shown), suggesting that sodin 5 and quinoin affect mycelial growth rather than conidial germination. As shown in Figures 2 and 6, sodin 5 and quinoin exhibited rRNA N-glycosylase activities against mammalian and fungal ribosomes, respectively. Thus, the antifungal activities of sodin 5 and quinoin against *P. digitatum* could be mediated by the inhibition of protein synthesis together with the ability to cross the membrane and enter into the fungal cells, as has been postulated for other type-1 RIPs [45,47].

**Figure 8.** Antifungal activity of sodin 5 (left panel) and quinoin (right panel) against *Penicillium digitatum*, measured in a microtiter plate bioassay. Conidia of *P. digitatum* were grown in Potato Dextrose Broth (PDB) for 24 h before exposure to different RIP concentrations. Fungal growth was followed for 70 h and measured as an increase in absorbance at 650 nm. The curves represent the buffer control or different amounts (μg/mL) of both toxins. The mean results ± SE of two experiments performed in triplicate are reported.

#### **3. Conclusions**

In conclusion, we have isolated seven type-1 RIPs from the different tissues of *S. soda* ('agretti' in Italian): five type-1 RIPs from seeds (sodins), one from edible leaves (sodin eL) and one from roots (sodin R). All these enzymes are able to release the β-fragment following incubation with rabbit or yeast ribosome and exhibit PNAG activity.

Sodin 5, the major form expressed in seeds (2.9 ± 0.15 mg/100 g of seeds), with respect to other type-1 RIPs from *S. soda* tissues, exhibits an α+β structure typical of type-1 RIPs with a high melting temperature (Tm = 76.03 ± 0.30 ◦C) and is non-glycosylated, as the other six sodins. Furthermore, sodin 5, sodin eL and sodin R show cytotoxic effects towards the HeLa and COLO 320 cell lines, inducing apoptosis. In addition, since fungi are among the most important plant pathogens, we tested the antifungal properties of both sodin 5 and quinoin against *P. digitatum*, finding that both RIPs possess concentration-dependent antifungal activity.

Overall, this research aims to revisit RIPs in edible plants in light of their possible use as antiviral, antifungal and antipathogenic tools in agri-food, overcoming the preconception about transgenic plants, as these enzymes are physiologically present in edible plants.

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

#### *4.1. Materials*

The chemicals for chromatography were previously reported [39,48,49]. Single-stranded salmon sperm DNA was obtained from Sigma-Aldrich (St. Louis, MO, USA). Quinoin from the seeds of *C. quinoa* and PD-L4 from the leaves of *P. dioica* were isolated as previously reported [34,39]. The nuclease-treated rabbit reticulocyte lysate system was purchased from Promega (Madison, WI, USA).

The medium and the other chemicals were from Sigma Chemical Co. (St. Louis, MO, USA). The RPMI 1640 medium, fetal bovine serum (FBS), penicillin, streptomycin and trypsin were purchased from GIBCO BRL (Barcelona, Spain). The Z-VAD-fmk (pan-caspase inhibitor carbobenzoxy-valyl-alanyl-aspartyl-[O-methyl]-fluoromethylketone) named Z-VAD was purchased from R&D Systems (Abingdon, UK).

Buffer A: 5 mM Na-phosphate, pH 7.2, containing 0.14 M NaCl; buffer B: 10 mM Na-acetate, pH 4.0; and buffer C: 5 mM Na-phosphate, pH 7.2.

#### *4.2. Purification of Type-1 RIPs from Seeds, Roots and Edible Leaves of S. soda*

Type-1 RIPs from *S. soda* were purified with the same protocol used for quinoin, type-1 RIP from *C. quinoa* seeds, as reported by Landi et al., 2021 [39]. Briefly, the crude extract in buffer A was first subjected to acid precipitation at pH 4.0 and cation step-wise chromatography using a SP-Streamline resin [column L × I.D. 20 cm × 30 mm, flow rate 3.0 mL/min; Cytiva, Buccinasco (MI) Italy]. Subsequently, the basic proteins, eluted with 1.0 M NaCl in buffer C, were gel-filtrated [HiLoad® 26/60 Superdex® column L × I.D. 60 cm × 26 mm, flow rate 2.5 mL/min (range 100–10 kDa); Cytiva] to separate the proteins by molecular weight, and then, basic proteins with a molecular weight of about 29 kDa were subjected to cation exchange chromatography on CM-Sepharose fast flow (Cytiva; column L × I.D. 25 cm × 16 mm) equilibrated in buffer C and eluted with a NaCl gradient up to 0.17 M (buffer C, 500 mL, buffer C containing 0.17 M NaCl, 500 mL; total volume 1 L using a peristaltic pump).

However, when the number of basic proteins after gel-filtration was lower (less than 200 μg), CM-Sepharose chromatography was replaced by FPLC on an AKTA Purifier System (Amersham Pharmacia; Milan, Italy) using a Source 15S PE 4.6/100 column, equilibrated in buffer C and eluted by a linear gradient from 0 to 50% of buffer C containing NaCl 0.3 M over 60 min (flow rate 1.0 mL/min). The same chromatographic step (Source 15S column) was carried out for minor forms of type-1 RIPs from the seeds of *S. soda* after cation exchange chromatography on the CM-Sepharose column.

#### *4.3. Enzymatic Assays*

#### 4.3.1. rRNA N-Glycosylase Activity of RIPs on Rabbit Ribosomes

The rRNA N-glycosylase assay was conducted as previously described [45]. Rabbit reticulocytes lysate (40 μL) was incubated with RIP (3.0 μg) at 37 ◦C for 1 h. After treatment, the RNA was extracted by phenolization, treated with 1 M aniline acetate (pH 4.5) and precipitated with cold ethanol. Purified RNA was analyzed by polyacrylamide gel in denaturing conditions [7 M urea/5% acrylamide (*w*/*v*)] and stained with ethidium bromide.

#### 4.3.2. rRNA N-Glycosylase Activity of RIPs on Yeast Ribosomes

The preparation of the 30,000 g (S30) supernatants from yeast was performed as described elsewhere [46]. The rRNA N-glycosylase activity was assayed in 50 μL samples of S30 supernatant from yeast, which was incubated with 5.0 μg of sodin 5, 0.7 μg of sodin eL, 1.5 μg of sodin R or 5.0 μg of quinoin for 1 h at 30 ◦C. After treatment, the RNA was extracted with phenol and treated with aniline for 10 min at 23 ◦C. The RNA samples were separated on a polyacrylamide gel in denaturing conditions [7 M urea/5% acrylamide (*w*/*v*)] and stained with Gel Red nucleic acid staining [50].

#### 4.3.3. Polynucleotide: Adenosine Glycosylase Activity on Salmon Sperm DNA

The adenine release was measured as previously reported [45], incubating salmon sperm DNA (10 μg) with RIPs (3.0 μg) in 300 μL 50 mM magnesium acetate (pH 4.0) containing 100 mM KCl, at 30 ◦C for 1 h. After incubation, the DNA was precipitated with cold ethanol and centrifuged. Adenine release was determined spectrophotometrically, reading the supernatant at 260 nm. On the other hand, to evaluate arbitrary units of PNAG activity on single fractions from *S. soda* seeds after CM-Sepharose chromatography, an equal volume was tested.

#### 4.3.4. Cell-Free Protein Synthesis Inhibition

The effect of RIPs on protein synthesis was determined through a coupled transcriptiontranslation in vitro assay using a rabbit reticulocytes lysate system, as described elsewhere [51]. Samples of RIPs were diluted and added to the reaction mixture as previously described [51]. Three experiments were conducted in duplicate, and IC50 (concentration that inhibits 50% protein synthesis) values were calculated by linear regression.

#### *4.4. Analytical Procedures*

The proteins' homogeneity was evaluated by SDS-PAGE with a Mini-Protean II (Bio-Rad; Milan, Italy) using a 6% stacking and 12% separating polyacrylamide gel under reducing conditions; a precision plus protein kit (Bio-Rad) was used as the reference proteins. The protein concentration was determined by a Pierce BCA Protein Assay Kit (Life Technologies Italia Fil., Monza, Italy). The glycosylation analysis was performed in gel after SDS-PAGE by using the Pro-Q™ Emerald 300 Glycoprot Probes Kombo (Life Technologies Italia). Glycosylated proteins were visualized by a ChemiDocTM XRS system.

#### *4.5. Circular Dichroism and Thermal Stability Determination*

The far-UV CD spectrum of sodin 5 was determined at 25 ◦C on a Jasco J-815 dichrograph [Jasco Europe, Cremella (LC) Italy]. A protein concentration of 0.15 mg/mL (5.15 μM) in 10 mM Na-phosphate, pH 7.2 (path-length quartz cuvette of 0.1 cm), was used for the far-UV spectrum measurements. DichroWeb (online analysis for protein Circular Dichroism spectra; http://dichroweb.cryst.bbk.ac.uk/html/home.shtml (accessed on 8 June 2022); [52]) was used to estimate the percentages of secondary structural elements.

Protein (~0.15 mg/mL) in 10 mM sodium phosphate, pH 7.2, was subjected to heatinduced denaturation, as previously reported [39].

#### *4.6. Cell Viability Assays*

The COLO 320 (human colon adenocarcinoma) and HeLa cell lines used in this study were obtained from the European Collection of Cell Cultures (ECACC). The cells were grown in RPMI 1640 medium (GIBCO BRL, Barcelona, Spain) supplemented with 10% fetal bovine serum (FBS), 100 U mL−<sup>1</sup> penicillin and 0.1 mg mL−<sup>1</sup> streptomycin under 5% CO2 at 37 ◦C. Cell viability was determined as previously reported [45]. The concentration of RIPs causing a 50% reduction in viability (IC50) was calculated by linear regression analysis. Sodin 5 and quinoin toxicity was also evaluated using HeLa cells pre-treated with 100 μM of the pan-caspase inhibitor Z-VAD. The reagent was added to cells 3 h before RIP administration, and the cell viability was determined for different RIP concentrations.

#### *4.7. DNA Fragmentation Analysis*

COLO 320 cells (1 × 106/plate) were incubated for 72 h in the presence of RIP (~0.5 μM). After treatment, cells were harvested by centrifugation (1000× *g* for 5 min). The pellets were lysed in 50 mM Tris Cl, pH 8.0, containing 10 mM EDTA and 0.5% SDS, and the DNA was isolated following the manufacturer's instructions [Genomic Prep Cells and Tissue DNA Isolation Kit (GE Healthcare, Madrid, Spain)]. DNA electrophoresis was carried out as previously reported [45].

#### *4.8. Antifungal Activity Measurements*

The growth inhibition assays of sodin 5 from *S. soda* seeds and quinoin from *C. quinoa* against *P. digitatum* were performed in 96-well microtiter plates. The conidia of *P. digitatum* (100 spores/well), obtained as indicated [47], were incubated at 26 ◦C in 150 μL PDB medium for 24 h to allow for conidia germination. The incubation was continued in the presence or in the absence of different RIP concentrations for a further 46 h. Fungal growth was followed for 70 h and measured as an increase in absorbance at 650 nm. Fungal growth was monitored spectrophotometrically using a microtiter plate reader (ELISA reader Multiskan) after 0, 24, 45, 56 and 70 h of incubation. The absorbance of cultures without cells was subtracted as the background.

**Supplementary Materials:** The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/toxins14080566/s1, Figure S1: SDS-PAGE analysis and in-gel staining for sugars of sodins isolated from *S. soda* tissues; Figure S2: Protein purification and polynucleotide:adenosine glycosylase activity of sodins 1–4; Figure S3: rRNA N-glycosylase activity of sodins from *S. soda* seeds assayed on rabbit ribosomes; Figure S4: staining for sugars of sodins isolated from *S. soda* tissues after SDS-PAGE.

**Author Contributions:** N.L., S.R., A.C. and H.Z.F.H. conducted the protein purification, characterization and enzymatic assays; L.C., R.I. and J.M.F. evaluated IC50 and performed cytotoxic activities; A.D.M. was responsible for the conceptualization, data analysis, writing and funding acquisition; S.R. and A.D.M. were responsible for the review and editing. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by the University of Campania 'Luigi Vanvitelli', by the project 'Svilupppo di Nutraceutici da Fonti Naturali—BIONUTRA', PON Ricerca e Innovazione 2014–2020 of the Campania region (code ARS01\_01166) and the Grant VA033G19 (Consejería de Educación, Junta de Castilla y León) to the GIR ProtIBio.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** The data presented in this study are available in this article.

**Acknowledgments:** The abnegation of all authors has made this study possible without dedicated funds considering the chronic difficulties afflicting the Italian research. The Authors are grateful for IT support to 'Maurizio Muselli', DiSTABiF.

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

#### **Abbreviations**

CD, circular dichroism; IC50, concentration that inhibits 50% of protein synthesis or reduces 50% of cell viability; PDB, potato dextrose broth; PNAG, polynucleotide:adenosine glycosylase; RIPs, ribosome-inactivating proteins; SRL, sarcin ricin loop; Tm, melting temperature.

#### **References**


## *Article* **Structure and Biological Properties of Ribosome-Inactivating Proteins and Lectins from Elder (***Sambucus nigra* **L.) Leaves**

**Rosario Iglesias 1,†, Rosita Russo 2,†, Nicola Landi 2, Mariangela Valletta 2, Angela Chambery 2, Antimo Di Maro 2, Andrea Bolognesi 3, José M. Ferreras 1,\* and Lucía Citores 1,\***


**Abstract:** Ribosome-inactivating proteins (RIPs) are a group of proteins with rRNA N-glycosylase activity that catalyze the removal of a specific adenine located in the sarcin–ricin loop of the large ribosomal RNA, which leads to the irreversible inhibition of protein synthesis and, consequently, cell death. The case of elderberry (*Sambucus nigra* L.) is unique, since more than 20 RIPs and related lectins have been isolated and characterized from the flowers, seeds, fruits, and bark of this plant. However, these kinds of proteins have never been isolated from elderberry leaves. In this work, we have purified RIPs and lectins from the leaves of this shrub, studying their main physicochemical characteristics, sequences, and biological properties. In elderberry leaves, we found one type 2 RIP and two related lectins that are specific for galactose, four type 2 RIPs that fail to agglutinate erythrocytes, and one type 1 RIP. Several of these proteins are homologous to others found elsewhere in the plant. The diversity of RIPs and lectins in the different elderberry tissues, and the different biological activities of these proteins, which have a high degree of homology with each other, constitute an excellent source of proteins that are of great interest in diagnostics, experimental therapy, and agriculture.

**Keywords:** anticancer agents; galactose; lectin; nanoLC–tandem mass spectrometry (nLC-MS/MS); protein synthesis (inhibition); ribosome-inactivating protein (RIP); ricin; sugar binding

**Key Contribution:** This work contributes to expanding our knowledge of the family of RIPs and RIP-related lectins produced by *Sambucus nigra*, with eight new proteins found in leaves: one type 2 RIP and two related lectins that are specific for galactose, four type 2 RIPs with deficient sugar binding domains and one type 1 RIP. This knowledge is important for the potential medical and biotechnological use of these proteins.

#### **1. Introduction**

Ribosome-inactivating proteins (RIPs) are a group of proteins with rRNA N-glycosylase activity (EC 3.2.2.22) that catalyze the elimination of a specific adenine located in the sarcin– ricin loop (SRL) present in the large rRNA of eukaryotes and prokaryotes [1,2]. The elimination of this adenine (A4324 in rat ribosomes, or the equivalent in other organisms) inactivates ribosomes, which leads to the irreversible inhibition of protein synthesis and, therefore, cell death [1,2]. RIPs have been classified according to their structure as type 1 RIPs, consisting of a polypeptide chain with N-glycosylase activity, and type 2 RIPs, formed by two polypeptide chains, an A (active) chain with enzymatic activity, and a B (binding) chain with lectin activity that can bind to receptors on the surface of cells, facilitating the entry of RIP [2]. Some type 2 RIPs, such as ricin, are extremely toxic, while others

**Citation:** Iglesias, R.; Russo, R.; Landi, N.; Valletta, M.; Chambery, A.; Di Maro, A.; Bolognesi, A.; Ferreras, J.M.; Citores, L. Structure and Biological Properties of Ribosome-Inactivating Proteins and Lectins from Elder (*Sambucus nigra* L.) Leaves. *Toxins* **2022**, *14*, 611. https://doi.org/10.3390/ toxins14090611

Received: 29 July 2022 Accepted: 29 August 2022 Published: 1 September 2022

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

**Copyright:** © 2022 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 (https:// creativecommons.org/licenses/by/ 4.0/).

have low toxicity; this is because the binding of the B chain to oligosaccharides present on the surface of cells is less effective, and because once internalized, the RIP follows an intracellular pathway different from ricin [1,3,4]. The toxicity of type 1 RIPs is lower, as they lack the lectin part and are, therefore, unable to bind to cells as type 2 RIPs do. Although the structure, activity, and mode of action of RIPs are known, their biological function is unclear. It has been proposed that these proteins could play an important role in the defense of plants against viruses, fungi, and insects [5,6].

Because of their diverse activities, RIPs, alone or as part of a conjugate, are good candidates for developing selective antiviral and anticancer agents [1,6]. Conjugates consist of a component directed against the target, such as an antibody, lectin, or growth factor, attached to a toxic component. RIPs have been used as the toxic component in several conjugates that have been tested in experimental therapies against various malignancies [3,7]. In agriculture, RIPs have been shown to increase resistance against viruses, fungi, and insects in transgenic plants [5,6].

RIPs are present in many angiosperm plants, both monocotyledonous and dicotyledonous, although in some plant families it is more common to find RIPs than in others; therefore, there are families such as Poaceae, Euphorbiaceae, Cucurbitaceae, Caryophyllaceae, Amaranthaceae, and Phytolaccaceae where several species with RIPs have been found, and other families where they have never been found [2]. Some species contain a wide variety of diverse RIPs, such as rice [8], or species of the genus *Phytolacca* [9,10].

The case of the genus *Sambucus*is unique, since more than 40 RIPs and lectins have been isolated and characterized to varying degrees from species belonging to this taxon [3,11]. Type 2 RIPs isolated from *Sambucus* have the peculiarity that, although they are enzymatically more active than ricin, they lack the high toxicity of ricin to cells and animals [11]. The presence in the same species of type 2 RIPs (heterodimeric and tetrameric), lectins (monomeric and homodimeric) structurally related to the above, together with type 1 RIPs, make the genus *Sambucus* an ideal model with which to study these proteins. Although these kinds of proteins can be found in other species, most of them have been obtained from elderberry (*Sambucus nigra* L.). Type 1 RIPs, type 2 RIPs (heterodimeric and tetrameric), and B-chain-related lectins have been obtained from the bark, seeds, flowers, and fruits, but the proteins from the leaves of this species have never been isolated. In this work we have isolated and characterized the most abundant RIPs and lectins in elderberry leaves by investigating their main physicochemical and structural properties, including amino acid sequence. We further studied their most important biological properties, and through in silico experiments, we explored potential mechanisms of sugar binding to these proteins.

#### **2. Results**

#### *2.1. Isolation of RIPs and Lectins from Elderberry Leaves*

Species of the genus *Sambucus* are one of the best sources for the isolation of RIPs and related lectins. These proteins have been found in *Sambucus ebulus* L. (dwarf elder), *S. nigra* L. (European elder), *S. sieboldiana* (Miq.) Blume ex Schwer. (Japanese elder), and *S. racemosa* L. (red elder) [3,11]. The most frequently used species, and the one with the greatest variety of these proteins, is *S. nigra*, in which RIPs and lectins have been isolated and characterized from bark, fruits, seeds, flowers, and pollen. Although RIPs from leaves of this species have not yet been isolated, in data banks the sequences of three proteins obtained from *S. nigra* leaves cDNA can be found: a type 2 RIP with an amino acid sequence similar to that of nigrin b from bark (named nigrin l), a monomeric lectin with a sequence similar to that of SNAIV from fruits (named SNAlm), and a homodimeric lectin with a sequence resembling that of SELld from the leaves of *S. ebulus* (named SNAld). Therefore, we aimed to isolate and characterize the RIPs and lectins from elderberry leaves. For this purpose, we optimized a standard RIP purification procedure [12] previously used to isolate RIPs and lectins from elderberry bark [13] and dwarf elder leaves [14]. A schematic overview of the procedures used to purify the RIPs and lectins from *S. nigra* leaves is shown in Figure 1.

**Figure 1.** Schematic overview of the procedures used to purify the RIPs and lectins from *S. nigra* leaves. All steps of purification are detailed in the Materials and Methods.

A great difficulty in carrying out this purpose is the extraordinary variability that these proteins present in the leaves. For this reason, two purifications were derived from the leaves of *S. nigra*. Two crude extracts were prepared from 700 and 540 g of leaves. Then, an acidified crude extract was obtained and subjected to ion exchange SP-Sepharose chromatography. After washing the column with sodium acetate (pH 4.5), the bound protein was eluted first with 5 mM sodium phosphate (pH 6.66), and then with sodium chloride. The fraction eluted with sodium phosphate displayed both protein synthesis inhibitory activity and erythrocyte agglutination ability, while the fraction eluted with sodium chloride inhibited protein synthesis but failed to agglutinate erythrocytes (data not shown). The yield of proteins was different in the two preparations: in the first, a higher amount of protein was obtained by eluting with sodium phosphate (1.4-fold higher); in the second, a better yield was obtained by eluting with sodium chloride (threefold higher). Therefore, we used the first preparation for purifying proteins eluted with sodium phosphate and the second preparation for purifying proteins eluted with NaCl.

The fraction eluted with sodium phosphate from the SP-Sepharose column was further purified by affinity chromatography using an acid-treated (AT)-Sepharose column. After washing with buffer, the D-galactose-binding proteins were eluted from the column with 0.2 M lactose, concentrated, and subjected to chromatography using a Superdex 75 HiLoad column (Figure 2a). The first peak contained nigrin l and SNAld, and the second contained SNAlm. The fractions of the first peak were dialyzed and subjected to Q-Sepharose chromatography. Proteins were eluted from the column with an NaCl gradient yielding two peaks (Figure 2b). The fractions of the first peak contained nigrin l and those of the second peak contained SNAld. The estimated yields of this preparation were 2, 12, and 1.4 mg per 100 g of leaves for nigrin l, SNAlm, and SNAld, respectively.

**Figure 2.** Purification of nigrin l, SNAlm, and SNAld: (**a**) The fraction eluted with lactose from the AT-Sepharose column was concentrated and chromatographed using Superdex 75 HiLoad. The fractions of the first peak contained nigrin l (Ngl) and SNAld, and those of the second peak contained SNAlm (horizontal bars); (**b**) the fractions of the first peak shown on panel (**a**) were dialyzed and subjected to chromatography using Q-Sepharose that was eluted with an NaCl gradient (dashed line), as indicated in the Materials and Methods. The fractions of the first peak contained nigrin l (Ngl) and those of the second peak contained SNAld (horizontal bars).

The fraction eluted with sodium chloride from the SP-Sepharose column was dialyzed and subjected to cation exchange chromatography using CM-Sepharose with a linear gradient of NaCl. As shown in Figure 3a, CM-Sepharose chromatography resolved several protein peaks. The fractions of the peaks were analyzed by electrophoresis in the presence and absence of 2-mercaptoethanol, and their effect on protein synthesis was tested. Further purification of the peaks yielded five proteins that strongly inhibited protein synthesis. Four of these new proteins corresponded to type 2 RIPs, and were named nigrin-Related Proteins 1–4 (nigrin-RPs 1–4). A new type 1 RIP was also found, and we named it nigritin l. For the purification of these new RIPs, the fractions derived via CM-Sepharose chromatography were collected as shown in Figure 3a, and subjected to subsequent chromatographies separately. Thus, nigrin-RP1 was purified by chromatography using Superdex 75 (Figure 3b). The protein obtained contained traces of nigrin l that were eliminated by chromatography using AT-Sepharose, as indicated in the Materials and Methods section. Nigrin-RP2 and nigrin-RP4 were also purified by chromatography using Superdex 75 (Figure 3c) and the trace contaminants were removed by re-chromatographing the proteins in the same column. To purify nigritin l, the fractions indicated in Figure 3a were subjected to chromatography using Superdex 75 (Figure 3d) and the contaminant traces were removed by re-chromatographing the protein on the same column. Finally, nigrin-RP3 was purified by chromatography using Superdex 75 (Figure 3e) followed by chromatography using SP-Sepharose with an NaCl gradient (Figure 3f). The estimated yields of the current preparation were 2.2, 2.6, 0.29, 0.03 and 1.5 mg per 100 g of leaves for nigrin-RPs 1–4 and nigritin l, respectively.

**Figure 3.** Purification of nigrin-Related Proteins 1–4, and nigritin l: (**a**) The proteins eluted with NaCl in SP-Sepharose were dialyzed and subjected to cation exchange chromatography using CM-Sepharose, as described in the Materials and Methods. The protein was eluted with a linear gradient of NaCl (dashed line). The fractions indicated by horizontal bars were separately pooled, concentrated, and subjected to molecular exclusion chromatography using Superdex 75 HiLoad; (**b**) Purification of nigrin-RP1. The fractions marked with NgRP1 in panel **a** were concentrated and subjected to chromatography using Superdex 75 HiLoad. The fractions indicated with the horizontal bar (NgRP1) were pooled, and the traces of nigrin l were eliminated by chromatography using AT-Sepharose, as indicated in the Materials and Methods; (**c**) Purification of nigrin-RP2 and nigrin-RP4. The fractions indicated with NgRP2/4 in panel (**a**) were concentrated and subjected to chromatography using Superdex 75 HiLoad. The fractions indicated with the horizontal bars (NgRP2 and NgRP4) were pooled and the contaminant traces were removed by re-chromatographing the proteins on the same column; (**d**) Purification of nigritin l. The fractions marked with Ngtl on panel (**a**) were concentrated and subjected to chromatography using Superdex 75 HiLoad. The fractions marked with the horizontal bar (Ngtl) were pooled and the contaminant traces were removed by re-chromatographing the proteins in the same column; (**e**,**f**) Purification of nigrin-RP3. The purification of nigrin-RP3 was conducted using the first preparation, with fractions equivalent to those indicated with NgRP3 in panel (**a**), which were concentrated and subjected to chromatography using Superdex 75 HiLoad (**e**). The fractions indicated with NgRP3 in panel **e** were dialyzed and subjected to chromatography using SP-Sepharose (**f**), which was eluted with an NaCl gradient (dashed line), as described in the Materials and Methods.

#### *2.2. Characterization of RIPs and Lectins from Elderberry Leaves*

Purified proteins from elderberry leaves were analyzed by electrophoresis on polyacrylamide gels in the presence of SDS (SDS-PAGE), and in the absence or presence of 2-mercaptoethanol. As shown in Figure 4, in the absence of a reductant, all of the proteins exhibited molecular weights between 50 and 65 kDa except for the lectin, SNAlm, and the type 1 RIP, nigritin l, which exhibited molecular weights of about 32.4 and 25.5 kDa, respectively. In the presence of 2-mercaptoethanol, all of the proteins produced bands with molecular weights between 25 and 35 kDa. Therefore, they are all dimeric proteins except

SNAlm and nigritin l, which are monomeric proteins. However, we found that reduction with 2-mercaptoethanol also induced changes in the apparent molecular weight of nigritin l, reducing the mobility of the protein to a molecular weight of about 27.5 kDa, suggesting the presence of intrachain disulfide bonds.

**Figure 4.** Analysis of purified proteins from elderberry leaves by electrophoresis on polyacrylamide gels: SDS-PAGE of the isolated proteins without (**a**) or with (**b**) 2-mercaptoethanol was carried out on 12% polyacrylamide separating gel and then stained with Coomassie brilliant blue. Samples of 5 μg of each protein were loaded on the gel except for nigritin l, for which 3 μg were loaded. The numbers indicate the corresponding size of the standards in kDa; (**c**,**d**) Sugar staining of RIPs and lectins from *S. nigra* leaves after SDS-PAGE. Each lane contained 5 μg of protein. SDS-PAGE was carried out on a 12% polyacrylamide separating gel in the presence of 2-mercaptoethanol, followed by Coomassie blue staining (**c**), or in-gel glycan detection (**d**) using the Pro-Q Emerald 300 glycoprotein staining kit. Stained glycoproteins were visualized by UV transillumination. Markers: CandyCane™ glycoproteins (**d**) and Coomassie blue (**c**) molecular weight standards.

The type 2 RIPs, nigrin l, and nigrin-RPs 1–4 are heterodimeric proteins consisting of a catalytic chain (A chain) and a lectin chain (B chain), both linked through a disulfide bond. In the presence of 2-mercaptoethanol, the apparent molecular weight values obtained for these proteins were 27.3 and 33.7 kDa for the two chains of nigrin l, 30.7 kDa and 33.7 kDa for the two chains of nigrin-RP2, and 27.9 kDa and 29.8 kDa for the two chains of nigrin-RP4. On the other hand, nigrin-RP1 is composed of two subunits of 30.6 kDa and nigrin-RP3 of two subunits of 30.2 kDa. The homodimeric lectin SNAld with an apparent molecular weight of 61.8 kDa contained only a homogeneous protein band of 30.9 kDa in the presence of 2-mercaptoethanol.

Many RIPs and lectins obtained from different species of *Sambucus* are glycoproteins; thus, we studied whether the purified proteins were glycosylated. Figure 4 compares proteins stained with Coomassie blue (Figure 4c) and those detected with a glycoprotein staining kit (Figure 4d). Quinoin (a glycosylated type 1 RIP) and PDL4 (a non-glycosylated type 1 RIP) are shown as controls. It can be observed that all the RIPs and lectins tested are strongly glycosylated, and, in those in which the A chain can be distinguished from the B chain, the latter is the most glycosylated. Thus, while the nigrin l B chain stained for carbohydrate, nigrin l A chain did not show any staining. On the contrary, nigrin-RP2 contained sugar chains on both subunits.

RIPs are potent inhibitors of protein synthesis in eukaryotes, as they are enzymes capable of inactivating ribosomes catalytically. Therefore, in mammalian cell-free systems, they usually show values of IC50 (concentration that inhibits 50% protein synthesis) in the ng/mL range. Accordingly, we tested all the proteins isolated from elderberry leaves in a coupled transcription–translation in vitro assay using a rabbit reticulocyte lysate system, finding the following values of IC50: 0.36, 0.75, 3.0, 0.35, 0.3 and 6.5 ng/mL for nigrin l, nigrin-RPs 1–4, and nigritin l, respectively (Figure S1). It is worth mentioning that, although all are good inhibitors of protein synthesis, there are great differences among the RIPs displaying IC50 values that differ up to 20 times. As expected, the lectins SNAlm and SNAld that lack a catalytic chain did not inhibit protein synthesis up to the maximum tested concentration of 1 μg/mL.

Type 2 RIPs consist of two chains, one being the enzymatic chain and the other being a lectin able to recognize sugars, mostly galactose residues. Due to the lectin activity, type 2 RIPs promote human erythrocyte agglutination. Among proteins isolated from *S. nigra* leaves, only nigrin l and the lectins SNAlm and SNAld can be purified by affinity chromatography using AT-Sepharose 6 B, which exposed galactose residues (Figure 1). The type 2 RIPs, nigrin-RPs 1–4, were not retained on AT-Sepharose (data not shown). We therefore studied the agglutination capacity of nigrin-RPs 1–4 and found that even a concentration of 200 μg/mL did not have any effect on human erythrocytes. Under the same conditions, nigrin l, SNALm, and SNALld agglutinated human erythrocytes at concentrations as low as 12.5, 40, and 6.2 μg /mL, respectively. Therefore, nigrin RPs 1–4, unlike nigrin l, may lack functional sugar binding domains. Similar type 2 RIPs have been previously described in *S. nigra* (SNLRPs 1 and 2) and *S. ebulus* (ebulin-RP) [14,15].

#### *2.3. rRNA N-Glycosylase, Adenine Polynucleotide Glycosylase, and DNA Nicking Activities*

RIPs are enzymes that irreversibly inactivate ribosomes because of their N-glycosylase activity (EC 3.2.2.22). The enzyme catalyzes the hydrolysis of the N-glycosidic bond between adenine number 4324 and its ribose in rat ribosomes (or equivalent adenine in sensitive ribosomes of other organisms) [1,2]. This activity can be evidenced by detecting, by means of a polyacrylamide gel electrophoresis, the RNA fragment released (Endo's fragment or diagnostic fragment) when the apurinic RNA is incubated in the presence of acid aniline [16]. As shown in Figure 5a, the type 2 RIPs, nigrin l and nigrin-RPs 1–4, and the type 1 RIP, nigritin l, cause, after treatment with acid aniline, the release of the diagnostic fragment of 460 nucleotides from rabbit reticulocyte ribosomes. As shown in Figure 5b, the type 2 RIPs can also depurinate yeast ribosomes, releasing a fragment of 368 nucleotides in the presence of acid aniline. Prokaryotic ribosomes are not sensitive to most RIPs. This is the case, for example, for ricin, volkensin [17], and other type 2 RIPs obtained from different species of the genus *Sambucus* [11]. However, they are sensitive to some type 1 RIPs, such as those obtained from *Pytolacca dioica* or *Beta vulgaris* [18,19]. As shown in Figure 5c, BE27 (obtained from the leaves of *B. vulgaris*) releases the diagnostic fragment from ribosomes of *Micrococcus lysodeikticus*, while the type 1 RIP, nigritin l, and the type 2 RIPs (Figure 5c and data not shown, respectively) from elderberry leaves do not, indicating that prokaryote ribosomes are not sensitive to these proteins.

**Figure 5.** Enzymatic activities of RIPs from *S. nigra* leaves: (**a**–**c**) rRNA N-glycosylase activity in animal, yeast, and bacterial ribosomes. The rRNA-glycosylase activity was tested as indicated in the Materials and Methods. Each lane contained 5 μg of RNA isolated from either untreated (control) or RIP-treated ribosomes from rabbit reticulocyte lysate (**a**), the yeast *Saccharomyces cerevisiae* (**b**), and 1 μg of RNA isolated from the bacterium *Micrococcus lysodeikticus* (**c**). The arrows indicate the RNA fragment (Endo's fragment) released as a result of the action of RIP after treatment with acid aniline (+). The numbers indicate the size of the markers in nucleotides; (**d**) Adenine polynucleotide glycosylase activity (APG) on DNA from salmon sperm. APG activity of 5 μg of RIP was assayed on salmon sperm DNA as described in the Materials and Methods, and the absorbance of the released adenine was measured at 260 nm. Data represent the mean of two duplicate experiments ± SE; (**e**) Nicking activity on pCR2.1 DNA. Samples comprising 200 ng/10 μL of plasmid DNA were incubated with 5 μg of RIP. Nigrin-RP1 (NgRP1) was also incubated in the presence of EDTA. R, L, and S indicate relaxed, linear, and supercoiled forms of pCR2.1, respectively. The numbers indicate the size of the markers in bp.

Some RIPs are also capable of removing more than one adenine from rRNA [20], and many of them can depurinate not only rRNA, but also other polynucleotide substrates, such as DNA, poly(A), mRNA, tRNA, and viral RNA. Therefore, the names adenine polynucleotide glycosylase (APG) and polynucleotide–adenosine glycosylase (PNAG) have been proposed for RIPs [21]. RIPs display different APG activities on DNA and RNA, all of which are capable of depurinating DNA from herring and salmon sperm [18,21]. However, this APG activity on DNA varies markedly between different RIPs. Generally, type 2 RIPs, such as ricin and kirkiin, have low activities [22], while some type 1 RIPs, such as those obtained from *Pytolacca dioica* or *Beta vulgaris*, possess very high activities [18,19]. Figure 5d shows the APG activities of RIPs from elderberry leaves on DNA compared with BE27 activity, measured as the absorbance at 260 nm produced by adenines released from salmon sperm DNA. As shown, RIPs from elderberry leaves have low activities, similar to ricin, stenodactylin, and kirkiin, compared to BE27 activity. Of these, the most active is nigrin l, which is twice as active as the others. However, none of the elderberry RIPs showed activity on tobacco mosaic virus RNA (data not shown).

Some RIPs exhibit topoisomerase (nicking) activity on plasmid DNA, transforming supercoiled DNA into the relaxed form; we tested whether elderberry RIPs possessed this activity, and found that only nigrin-RP1 and the type 1 RIP, nigritin l, were able to convert the supercoiled PCR 2.1 DNA forms into the relaxed forms (Figure 5e). Such activity was dependent of Mg2+ ions because it was inhibited by EDTA (Figure 5e).

Therefore, elderberry leaves contain mainly a type 1 RIP (nigritin l), which displayed rRNA N-glycosylase activity, a type 2 RIP with N-glycosylase activity and lectin activity (nigrin l), four type 2 RIPs with N-glycosylase activity that fail to agglutinate erythrocytes (nigrin-RPs 1–4), and a monomeric (SNAlm) and a dimeric lectin (SNAld).

#### *2.4. Peptide Mapping of RIPs from S. nigra Leaves by High-Resolution MS/MS*

The proteins isolated from *S. nigra* leaves, nigrin l, nigrin-RPs 1–3, SNALm, and SNAld were characterized via a peptide mapping approach based on high-resolution nanoLC– tandem mass spectrometry (Figure 6). Nigrin-RP4 was not studied due to the low levels obtained via the purification process. Similarly, nigritin l was not further investigated due to the lack of sequence data available for database searches. Preliminary optimization of sample preparation steps for MS analyses was performed to define the conditions for the reduction, alkylation, and tryptic hydrolysis of *S. nigra* proteins. An extensive step of protein denaturation and disulfide bridge reduction performed with 20 mM DTT at 95 ◦C was needed to obtain a significant yield of tryptic peptides suitable for MS analysis. Then, free cysteinyl residue alkylation with iodoacetamide (IAA) and enzymatic proteolysis with trypsin were both performed before nanoLC–ESI–MS/MS analyses on a Q Exactive Orbitrap mass spectrometer. A data-dependent acquisition mode was used, during which higher-energy collisional dissociation (HCD) MS/MS spectra were obtained for the five most intense mass peaks in each scan, allowing for accurate amino acid sequencing of tryptic peptides. By this approach, for the monomeric lectin, 18 peptide spectral matches (PSMs) were mapped on the SNAlm sequence (AC: AAN86132). Amino acid sequences of peptides obtained by high-resolution tandem mass spectrometry are reported in Table S1. Representative MS/MS spectra of peptides are reported in Figure S2. Regarding the dimeric lectin, nine PSMs were mapped on the SNAld sequence (AC: AAN86131, Table S2 and Figure S3).

**Figure 6.** Schematic overview of the experimental workflow used for the peptide mapping of proteins from *Sambucus nigra* leaves by high-resolution nanoLC–MS/MS.

A high number of MS/MS spectra (53 PSMs) of the type 2 RIP, nigrin l, were mapped on the A chain and B chain of nigrin l (AC: AAN86130, Table S3 and Figure S4).

For both nigrin-RP 1 and 2, sequenced peptides were mapped on SNLRP2 A and B chains (AC: AAC49672, Tables S4 and S5) (23 and 24 PSMs, respectively). On the contrary, nigrin-RP3 was identified as SNLRP1 (AC: AAC49673, Table S6) (27 PSMs). Representative MS/MS spectra of peptides from nigrin-RPs 1, 2, and 3 are reported in Figures S5–S7.

Peptide mapping of proteins purified from *S. nigra* leaves by high-resolution MS/MS identifies the proteins obtained by affinity chromatography using AT-Sepharose with three sequences obtained from *S. nigra* leaves cDNA: the type 2 RIP, nigrin l, and the lectins SNAlm and SNAld (accession numbers AAN86130, AAN86132, and AAN86131, respectively). On the other hand, nigrin-RP3 would be a type 2 RIP homologous to SNLRP1 from elderberry bark (accession number AAC49673), and nigrin-RPs 1 and 2 would be homologous to the type 2 RIP, SNLRP2, also found in the bark of elderberry (accession number AAC49672). Alternatively, nigrin-RPs 1 and 2 could be the same protein with different degrees of glycosylation (see Figure 4d). Evidence against this hypothesis is the fact that these two proteins exhibit different protein synthesis inhibitory activities (IC50 fourfold higher for nigrin-RP2) and different behaviors against the supercoiled plasmid (see Figure 5e). Figure 7 graphically presents these data by comparing the sequences obtained by mass spectrometry with the sequences found in the data banks. It is noteworthy that neither SNLRP1 nor SNLRP2 agglutinate erythrocytes [15].

#### *2.5. Carbohydrate Binding Properties of Nigrin l, SNAlm, and SNAld*

Nigrin l, SNAlm, and SNAld showed hemagglutination activities in human erythrocytes (Section 2.2). To elucidate the sugar binding specificities of these proteins, hemagglutination inhibition with various monosaccharides and disaccharides was carried out (Table 1). The results show that the agglutination produced by the three proteins was inhibited by D-galactose and lactose (β-D-galactopyranosyl-(1→4)-D-glucose). In none of the three proteins was an affinity for D-glucose, D-fructose, D-mannose, or L-fucose observed at the maximum sugar concentration tested. The protein showing the highest affinity for galactose was SNAlm, whereas SNAld showed the lowest affinity. In all cases, lactose inhibited agglutination at a concentration four times lower than galactose.


**Figure 7.** Comparison of the sequences of RIPs and lectins from leaves of *S. nigra* obtained by peptide mapping via high-resolution nanoLC–MS/MS mass spectrometry with sequences deposited in the data banks. The gray shaded sequences were obtained by peptide mapping via high-resolution MS/MS mass spectrometry and match sequences obtained from the data banks with the access numbers AAN86130 (nigrin l), AAN86132 (SNAlm), AAN86131 (SNAld), AAC49672 (nigrin-RP1 and 2), and AAC49673 (nigrin-RP3). The red boxes indicate the amino acids that possibly form the catalytic pocket, and the blue boxes indicate the amino acids that possibly form the 1α and 2γ sugar binding sites. Identical residues (\*), conserved substitutions (:), and semiconserved substitutions (.) are reported.


**Table 1.** Inhibition of the hemagglutination activity of nigrin l, SNAlm, and SNAld by sugars.

<sup>1</sup> No inhibition of hemagglutination at the maximum sugar concentration tested (200 mM) was observed with the following sugars: D-glucose, D-fructose, D-mannose, and L-fucose.

#### *2.6. Structural Analysis of Nigrin l, SNAlm, and SNAld*

Given the availability of the complete amino acid sequence, it was possible to predict the three-dimensional structures of nigrin l, SNAlm, and SNAld with a computational model using the potentials of deep learning and neural networks [23]. The best threedimensional models obtained for nigrin l, SNAlm, and SNAld are shown in Figure 8, and all showed local Distance Difference Test (lDTT) values that were much higher than 90%, which makes them suitable for characterizing the binding sites [23]. As described for ricin and other type 2 RIPs [24,25], the A chain of nigrin l consists of three folding domains. The first domain includes the N-terminal, and is composed of six antiparallel β-sheets and two α-helices in the order aAbcdeBf. The second domain consists of five α-helices (helices from C to G). The last domain consists of two α-helices and two antiparallel β-sheets in a α-helix–β-fork–α-helix motif (HghI). Similar to other type 2 RIPs, the B chain is made up of two homologous globular lectin domains arising from gene duplication, which are made up exclusively of β-sheets. Each domain consists of four homologous subdomains (1λ, 1α, 1β, and 1γ for lectin 1; 2λ, 2α, 2β, and 2γ for lectin 2). The subdomains 1λ and 2λ are responsible for the linking to the A chain and the interconnection between the two domains of the B chain, respectively. The subdomains 1α, 1β, and 1γ are arranged in a trefoil structure. This arrangement is also present in lectin 2 with the subdomains 2α, 2β, and 2γ. The 1α and 2γ subdomains contain the two D-galactose binding sites. SNAlm and SNAld have structures similar to nigrin l, but both lack the A chain. In addition, SNAld has an additional cysteine (Cys 23) that allows it to form a dimer with another identical chain.

**Figure 8.** Comparison of the structures of ricin, nigrin l, SNAlm, and the monomer of SNAld. The three-dimensional structural modeling was carried out using AlphaFold2 software, and the figure was generated using Discovery Studio 2021. The arrows indicate the position of the disulfide bond linking A and B chains.

#### *2.7. Molecular Docking*

The availability of 3D models of the proteins that allow studies at the molecular level encouraged us to study how D-galactose binds to the 1α and 2γ sites of nigrin l, SNAlm, and SNAld. For this purpose, we carried out a molecular docking study using Autodock 4.2, and compared the results with those already published for ricin [26].

As shown in Figure 9, the sequences of the binding sites for sugars of nigrin l, SNAlm, and SNAld are similar to those of ricin. Seven of the fourteen amino acids in the binding pocket of the 1α site are identical, as are seven out of twelve amino acids at the 2γ site. This is consistent with the fact that all these proteins are specific for D-galactose and lactose. However, there are differences with the sugar binding sites of ricin that may influence the affinity of these sites for sugars. The 1α site is identical in nigrin l and SNAlm. In addition, these sites are relatively similar to that of ricin. In all three proteins, the binding of β-D-galactopyranose is the result of the C–H–π interaction between the aromatic rings of tryptophan (W37, W39, and W33 in ricin, nigrin l, and SNAlm, respectively) and the apolar face of the pyranosic ring of galactose. The C–H groups of carbons 3, 4, 5, and 6 are oriented towards the aromatic rings of tryptophan, allowing the π interaction of the electron cloud with the aliphatic protons of sugar that carry a positive partial charge. The polar face of galactose forms hydrogen bonds with five amino acids located on the other side of the binding pocket, four of which (aspartic, arginine, glutamine, and asparagine) are identical in all three proteins. The 1α site of the SNAld is very different. This is likely influenced by the presence within this site of the cysteine (C23) that forms the disulfide bond with the other subunit. Nevertheless, both galactose and lactose can bind at the 1α site of SNAld, although galactose adopts a different arrangement that can affect the affinity for this sugar. In this case, carbons 1, 5, and 6 are oriented towards the aromatic rings of tryptophan, and only coincide, on the other side, the amino acids aspartic and arginine.

At the 2γ site, the amino acids that provide aromatic rings are tyrosine (in ricin and SNAlm) and phenylalanine (in nigrin l and SNAld). However, these residues seem to not affect the orientation of the pyranosic ring at this site, as it is virtually identical. The C–H groups of carbons 4 and 6 are oriented towards the aromatic ring, allowing the interaction with the apolar face of D-galactose, while the polar face forms hydrogen bonds with various amino acids, of which at least four are identical in the three proteins. In ricin, the orientation of galactose is slightly different since the C–H groups of carbons 3, 4, and 5 are oriented towards the aromatic ring, while on the other side, only two amino acids (aspartic and asparagine) coincide.

Similar to D-galactose, lactose can also bind to the 1α and 2γ sites of nigrin l, SNAlm, and SNAld (Figures S8 and S9). However, in all cases, lactose inhibited the agglutination of erythrocytes at a concentration four times lower than D-galactose (Table 1). This is not in accordance with the estimated free energy of binding data provided by Autodock 4.2 (Table S7). One possible explanation is that lactose contains D-galactose in the β-pyranosic form, allowing effective binding to aromatic amino acids at sugar binding sites.

It is worth mentioning that peptide mapping identifies the sequences of the binding sites of nigrin-RPs 1–3 (Figure 7), which match with the sequences of SNLRP1 and SNLRP2 [15]. These are proteins from the bark of *S. nigra* that, similar to nigrin-RPs, do not agglutinate erythrocytes. This was initially attributed to the fact that the 1α and 2γ sites were inactive because amino acid substitutions at these sites prevented carbohydrate binding [15]; however, SNLRP has subsequently been reported to interact with N-acetylglucosamine oligomers, as well as with many glycan structures containing Nacetylglucosamine [27].

#### **Figure 9.** Three-dimensional models of the galactose binding sites of ricin, nigrin l, SNAlm, and SNAld. The alignment of the sequences of subdomains 1α and 2γ of ricin, nigrin l, SNAlm, and SNAld are represented. The binding pockets of sites 1α and 2γ are indicated by blue boxes. Amino acids involved in galactose binding by C–H–π interactions or hydrogen bonds are colored purple and cyan, respectively. Identical residues (\*), conserved substitutions (:), and semiconserved substitutions (.) are reported. The galactose binding sites of ricin (PDB 2AAI), nigrin l, SNAlm, and SNAld complexed with β-D-galactopyranose (thick sticks) are represented. The amino acids that bind the galactose molecule, either by C–H–π interactions or both conventional and nonconventional hydrogen bonds, are represented by thin sticks. At the bottom, the positions of the centroids of the pyranosic ring of D-galactose with respect to the aromatic rings of sites 1α and 2γ of ricin, nigrin l, SNAlm, and SNAld are represented by green balls.

#### *2.8. Cytotoxic Effect of RIPs from Elderberry Leaves in Cell Cultures*

Type 1 RIPs consisting of a single enzymatic (A) chain usually display lower toxicity than type 2 RIPs consisting of a binding (B) chain with lectin activity linked to the enzymatic A chain. The carbohydrate binding domains of the B chain recognize glycosylated receptors on the cell surface, facilitating the entry of the A chain into the cell. Ricin is a wellknown example of a highly toxic type 2 RIP. However, all type 2 RIPs found in the genus *Sambucus* are considered nontoxic type 2 RIPs since, despite the high enzymatic activity on ribosomes comparable to that of ricin, they show much lower toxicity to cells and animals [3,4,11]. Figure 10a shows the toxicities of RIPs from *S. nigra* leaves towards HeLa and COLO 320 cells. The RIPs from *S. nigra* leaves were toxic to HeLa and COLO 320 cells, exhibiting IC50 (concentration of protein causing the death of 50% cells) values ranging from 19 to >1460 nM. In all cases, the cytotoxicity of these RIPs was much less than that exerted by ricin, which affects viability, with IC50 values several orders of magnitude lower (0.14–0.6 pM) [14]. The most sensitive were HeLa cells, showing IC50 values ranging from 19 to 580 nM, while COLO 320 cells exhibited values between 19 and >1460 nM after 48 h of treatment. When comparing the type 2 RIPs nigrin l, a galactose-binding protein, and nigrin-RPs 1–3, nigrin l was the most active toxin, with IC50 values of 19 nM for COLO 320 and HeLa cells. Nigrin-RPs 1–3 displayed very low toxicity, especially on COLO 320 cells, comparable to that of the type 1 RIP, nigritin l. The IC50 values obtained from treated HeLa cells were 130, 340, 580, and 280 nM for nigrin-RPs 1–3 and nigritin l, respectively. The lack of sugar binding activity of nigrin-RPs could play a role in their low cytotoxicity. HeLa cells treated with nigrin l, nigrin-RPs, and nigritin l exhibited morphological features characteristic of apoptosis, such as cell rounding and blebbing (data not shown). Several studies reported that the cytotoxicity of RIPs is associated, in addition to rRNA damage, with their ability to induce apoptosis [28]. To investigate the capability of nigrin l to reach the cytosol and inactivate the ribosomes after being endocytosed, we analyzed the ribosomal RNA from HeLa cells treated with the RIP for 48 h. Figure 10b shows that the ribosomes were depurinated, releasing the diagnostic fragment after treatment of the RNA with acid aniline, indicating that nigrin l was able to reach the ribosomes to inhibit protein synthesis. Apoptosis might be a consequence of the ribotoxic stress induced by the RIP after entry into the cytosol, or both processes could run in parallel. To determine whether the observed cytotoxic effects of nigrin l were also mediated via apoptosis in COLO 320 cells, we evaluated the breakdown of the nuclear DNA into oligonucleosomal fragments. As shown in Figure 10c, when COLO cells were treated with 40 nM nigrin l for 72 h, the cleavage of the chromosomal DNA was clearly observed. Thus, our data suggest that the apoptotic pathway was involved in the cell death mediated by RIPs from *S. nigra* leaves.

**Figure 10.** Cytotoxicity of nigrin l, nigrin-RPs 1–3, and nigritin l on HeLa and COLO 320 cells: (**a**) Effect of nigrin l (•), nigrin-RP1 (-), nigrin-RP2 (), nigrin-RP3 (), and nigritin l () on viability of COLO 320 (left panel) and HeLa (right panel) cells. Cells were incubated with different concentrations of RIPs for 48 h, and cell viability was evaluated by a colorimetric assay, as indicated in the Materials and Methods. Data represent the mean ± SD of two experiments performed in duplicate; (**b**) N-glycosylase activity of nigrin l on rRNA from HeLa cells. rRNA N-glycosylase activity was assayed as indicated in the Materials and Methods. Each lane contained 2 μg of RNA isolated from either untreated cells (C, control) or cells incubated with 40 nM of nigrin l for 48 h. The arrow indicates the RNA fragment released as a result of RIP action upon acid aniline treatment. Numbers indicate the size of the standards (M) in nucleotides; (**c**) Effect of nigrin l on internucleosomal DNA fragmentation. COLO 320 cells were incubated in the absence (C, control) or presence of 40 nM of nigrin l for 72 h. The DNA was isolated and 4.0 μg was electrophoresed, as indicated in Section 5.3.15. The numbers indicate the corresponding size of the standards (M) (λDNA HindIII/EcoRI) in pb.

#### **3. Discussion**

Species of the genus *Sambucus* are one of the best sources of ribosome-inactivating proteins. From different tissues, type 1 RIPs, type 2 RIPs (heterodimeric and tetrameric), and lectins (monomeric and dimeric) related to the B chain of type 2 RIPs have been obtained [3,11]. Type 2 RIPs isolated from *Sambucus* are peculiar in that they lack the toxicity of other type 2 RIPs, such as ricin [29], abrin [30], stenodactylin [28], and kirkiin [22]. For example, nigrin b is 50,000 times less toxic than ricin to HeLa cells, and 1,500 times less toxic to mice [31]. Therefore, they have been used as the enzymatic component of immunotoxins and other conjugates directed against tumor cells [3,11]. Due to their specificity for the α2,6-linked sialic acid, SNAI from the bark of *S. nigra* and SSA from the bark of *S. sieboldiana* are used in highly diverse biomedical applications, such as diagnosis by ELISA [32,33], histochemistry [34,35], confocal fluorescence microscopy [36], microarrays [37], and new therapeutic strategies [38]. Elderberry RIPs have also been used in agriculture to obtain transgenic plants resistant to viruses [39,40] and insects [41,42].

Most of these proteins have been obtained from *S. nigra*. Thus, from the bark of this species, a galactose-specific type 2 RIP (nigrin b or SNAV) [43,44]; a sialic-acid-specific type 2 RIP (SNAI') [45]; three type 2 RIPs that do not agglutinate erythrocytes, and that could be specific for N-acetylglucosamine (SNLRP1, SNLRP2, and basic nigrin b) [15,27,46]; a sialic-acid-specific tetrameric type 2 RIP (SNAI) [47]; and a galactose-specific monomeric lectin (SNAII) [48] have been purified. From the fruits two type 1 RIPs (nigritins f1 and f2) [49], a galactose-specific type 2 RIP (nigrin f) [50], a sialic-acid-specific tetrameric type 2 RIP (SNAIf) [51], and a galactose-specific monomeric lectin (SNAIV or SNAIVf) have been obtained [52]. From the seeds, nigrin s (type 2 RIP) [53] and SNAIII (monomeric lectin) [54], both specific for galactose, have been purified. Finally, the presence of SNAflu-I, a tetrameric type 2 RIP reported as being specific for galactose, has been described in the flowers of *S. nigra* [55,56]. Some of them can be considered isoforms; for example, SNAI from the bark and SNAIf from fruits have amino acid identities of 95%, and can therefore be considered as tissue-specific isoenzymes.

However, even though elderberry has been the subject of intense research for the search and study of RIPs and lectins since 1984 [57], leaf proteins have never been purified, despite crude leaf extracts exhibiting very potent activities in both protein synthesis inhibition in rabbit reticulocyte lysate (IC50 = 50 ng/mL) and hemagglutination (minimum concentration agglutinating erythrocytes = 1 mg/mL) (data not shown). One of the reasons for this is the difficulty of isolating a considerable number of proteins with similar characteristics, and which also present great developmental variations in their expression in this tissue (data not shown). In spite of this, we proposed to investigate the presence of RIPs and lectins in the leaves in order to isolate and characterize them to enhance our knowledge of these proteins.

We found three proteins that are specific for galactose: nigrin l (a heterodimeric type 2 RIP with an A chain of 27.3 kDa and a B chain of 33.7 kDa), SNAlm (a 32.4 kDa monomeric lectin), and SNAld (a homodimeric lectin with two identical 30.9 kDa subunits) (Figure 4a,b). These proteins could be identified by peptide mapping with three sequences obtained from *S. nigra* leaves cDNA. The sequenced peptides matched with the sequences with access numbers AAN86130 (nigrin l), AAN86132 (SNAlm), and AAN86131 (SNAld), with coverages of 59, 53, and 38%, respectively (Figure 7). Nigrin l can be considered an isoenzyme of nigrin b from the bark since they have an amino acid identity of 98.4%. Likewise, SNAlm can be considered an isoform of SNAIV (or SNAIVf) from fruits, with which it shares 90.2% of the amino acids. However, no dimeric lectins have been found in other elderberry tissues; thus, the most related protein is SELld, found in the leaves of *S. ebulus* [58], with which it presents a homology of 89.9%. Therefore, these dimeric lectins appear to be unique to the leaves. It is also noteworthy that in leaves, no tetrameric type 2 RIPs corresponding to SNAI from the bark, SNAIf from the fruits, or SNAflu-I from flowers were found, tissues in which this type of structure is among the predominant [13,51,55–57].

We also found four heterodimeric type 2 RIPs that fail to agglutinate erythrocytes (nigrin-RPs 1, 2, 3, and 4). Nigrin-RPs 1 and 3 are heterodimers whose A and B subunits have a similar molecular weight of about 30 kDa, whereas nigrin-RPs 2 and 4 are heterodimers whose A and B subunits have molecular weights of 30.7 and 33.7 kDa for nigrin-RP2 and 27.9 and 29.8 kDa for nigrin-RP4 (Figure 4a,b). All these proteins are strongly glycosylated, a characteristic they share with most RIPs and lectins from the genus *Sambucus* [14,46,49,50,59,60]. Peptide mapping identified nigrin-RP3 as an isoform of SNLRP1 from the bark, and nigrins-RPs 1 and 2 as SNLRP2 isoforms (Figure 7). Unfortunately, not enough nigrin-RP4 was obtained to perform peptide mapping. The data indicate that nigrin-RPs 1 and 2 could be isoforms with different amino acid sequences or proteins with the same sequence, but different states of glycosylation. In favor of the latter hypothesis is the fact that no differences in the amino acid sequences were found in peptide mapping (Figure 7). Moreover, the B chain of nigrin-RP2 is strongly glycosylated (Figure 4d), suggesting the occurrence of a different glycosylation pattern for the B chain of nigrin-RP2 compared to that of nigrin-RP1 (Figure 4d). This could also explain the different

behaviors of the two proteins when interacting with the CM-Sepharose chromatography column (Figure 3a) and in SDS-PAGE (Figure 4). However, the evidence that the two proteins display different enzymatic activities, with nigrin-RP1 being more active, both as an inhibitor of protein synthesis and in nicking activity, leads us to hypothesize that they are probably proteins with different amino acid sequences. Finally, we also found a type 1 RIP, nigritin l, which we have not been able to map. However, based on its behavior in chromatography columns, electrophoresis, and its enzymatic activities (for example, its nicking activity), it could very likely correspond to an isoform of nigritin f1, a protein isolated from fruits of *S. nigra* [49].

One of the most important features of RIPs to consider is their enzymatic properties, both for their possible biological role and for their biotechnological applications. RIPs are inhibitors of protein synthesis using rRNA N-glycosylase activity, which catalyzes the elimination of a specific adenine located in the sarcin–ricin loop (SRL) that is present in the large rRNA of eukaryotes and prokaryotes [1,2]. All the RIPs from elderberry leaves were shown as strong inhibitors of protein synthesis in rabbit reticulocyte lysate, being the most potent nigrin l, nigrin-RP3 and nigrin-RP4 which presented an IC50 20-fold lower than nigritin l. Since the sequence of the active site is very similar in all proteins (Figure 7), the difference could be attributed to small differences in the active site, but mainly to differences in the amino acid sequence of the A chain that binds to ribosomal proteins [61,62]. All RIPs in this study showed rRNA N-glycosylase activity, not only in rabbit reticulocyte ribosomes, but also in yeast ribosomes (Figure 5a,b). However, unlike BE27, none showed activity against bacterial ribosomes (Figure 5c). This is often considered an advantage for biotechnological applications because it facilitates their cloning and expression in bacteria.

Although RIPs are classified as rRNA N-glycosylases, one very important enzymatic activity of these proteins is the depurination of nucleic acids (adenine polynucleotide glycosylase or APG activity) [21]. Different RIPs show different APG activities on DNA and RNA; however, all of them can depurinate the DNA of herring and/or salmon sperm, and some of them can also depurinate various types of RNA [18,19,21]. Although this activity has only been demonstrated in vitro, it could be important for apoptotic activity against animal cells or antiviral activity [19]. All RIPs tested showed APG activity on salmon sperm DNA (Figure 5d), similar to that of ricin and kirkiin [22], although much lower than that of BE27 [19], since type 1 RIPs usually have an APG activity on DNA 10-fold higher than type 2 RIPs [21]. Of all the RIPs in elderberry leaves, the one that displayed the highest activity was nigrin l, which showed a twofold higher activity with respect to the other proteins.

Some RIPs show endonuclease (nicking) activity on the DNA of supercoiled plasmids that produces relaxed and sometimes linear plasmids. This ability may be necessary for these proteins to perform different biological functions, including resistance to pathogenic microorganisms or viruses [18,19]. Only nigrin-RP1 and nigritin l promoted the conversion of supercoiled DNA into relaxed forms. This activity was dependent on magnesium ions, as it was inhibited by EDTA, according to what has been reported for other RIPs [18,19]. Regarding this activity, nigrin-RP1 and nigrin-RP2 showed very different behaviors, so the differences between both proteins must be more than just a difference in the glycosylation pattern. Many RIPs do not present this activity, which is related to a different configuration in the structure near the active site that allows the accommodation of the supercoiled DNA of the plasmids [19,63].

Nigrin l, SNAlm, and SNAld were revealed to be lectins with affinities for galactose (Table 1), the affinity of SNAlm for both galactose and lactose being significantly higher. This contrasts with some elderberry RIPs that have affinities for sialic acid, and appear not to be present in elderberry leaves. Due to the arrangement of its hydroxyl groups, β-D-galactopyranose has two faces, a polar face and an apolar face. In several RIPs such as ricin, abrin, stenodactylin, and kirkiin, the binding of β-D-galactopyranose to the 1α and 2γ sites is the result of the C–H–π interaction between the aromatic ring(s) of an amino acid (tryptophan, phenylalanine, tyrosine, or histidine) and the apolar face of the pyranosic

ring of galactose, and hydrogen bonds between the hydroxyl groups of the polar face and amino acids located on the other side of the pocket of binding [24,25].

The recent emergence of programs that predict the structure of proteins at the atomic level [23,64] has allowed us to conduct in silico experiments to predict how galactose and lactose bind to these proteins. A surprising finding is that the 1α site of the SNAld is fully functional, although the subdomain in which it is located is also used for the linking of the two monomers that form the dimeric protein (Figure 9). In SNAld, the 1α site has undergone numerous changes, including the deletion of five amino acids, which allows it to keep away the 1α site of the other subunit without affecting the galactose binding capacity (Figure S10). The way galactose binds to the 1α site of nigrin l and SNAlm is practically identical, and very similar to that of ricin, although some of the amino acids with which it forms hydrogen bonds change, which could affect the affinity for sugars (Figure 9). However, the orientation of galactose at the 1α site of SNAld is different, caused by what we have discussed above, which does not seem to affect the affinity for monosaccharide with respect to the other two proteins. In the case of the 2γ site, the way galactose binds to the three elderberry proteins is practically identical, and somewhat different from that of ricin (Figure 9). It has been suggested that the difference between toxic RIPs, such as ricin, and nontoxic RIPs such as ebulin l or nigrin b, lies in the 2γ site, which binds galactose in a different way, resulting in a decreased affinity for membrane glycoproteins containing galactose and, as a consequence of the deficient binding, an intracellular pathway different from that of ricin [3,11]. The most notable change at this site is the substitution of tyrosine for phenylalanine, which seemed to suggest that this change alone could modify the way galactose binds. In the case of ebulin l, this prevents the binding of lactose, which explains the fact that this protein has less affinity for the α-lactose-agarose gel [65]. Nigrin b, which has a 98.4% homology with nigrin l, also has less affinity for the AT-Sepharose matrix [13]; however, we have not observed that the binding of galactose to the 2γ site of nigrin l is affected by glucose as part of lactose (Figure S9). Nor does the change from tyrosine to phenylalanine seem to be decisive because SNAlm contains tyrosine and does not exhibit a different mode of galactose binding from that of nigrin l. However, this different affinity could be explained by the different way galactose binds to this site with respect to ricin (Figure 9).

Peptide mapping identified the sequence of the binding sites of nigrin-RPs l, 2, and 3 with the sequences of SNLRP1 and SNLRP2, type 2 RIPs that fail to agglutinate erythrocytes and do not recognize galactose [15]. In these proteins, the most prominent change occurs at the 1α site, which lacks an aromatic amino acid. The 2γ site contains tyrosine; however, several changes occur in the amino acids that form hydrogen bonds with galactose, which could affect the monosaccharide binding to this site. On the other hand, ebulin-RP is a heterodimeric type 2 RIP present in *S. ebulus* leaves that displays rRNA N-glycosylase activity but lacks functional sugar binding domains [14]. Moreover, the B chain of ebulin-RP shares a strong homology (83.14%) with the B chains of SNLRPs, and thus with those of nigrin-RPs. It has been proposed, based on molecular docking, that the loss of lectin activity of ebulin-RP may be due to the presence of inactive 1α and 2γ sites, and this could also be the case for nigrin-RPs from leaves.

We have performed a study of the cytotoxic activity of *S. nigra* leaf RIPs, including the type 2 RIPs (nigrin l and nigrin-RPs 1–3) and the type 1 nigritin l, towards COLO 320 and HeLa cells. First, our data confirmed that the elderberry type 2 RIPs are much less toxic than ricin, which affects the viability of these cells, with IC50 five–six orders of magnitude lower [14]. The reason for the different toxicities among type 2 RIPs is not clear. Type 2 RIPs from *Sambucus* leaves, inactivate ribosomes in vitro with higher efficiency than ricin, therefore, the different toxicities could be better attributed to differences between the B chains, which are responsible for the interaction with cellular membranes, than to the enzymatic A chains. In this line, it has been shown that the type 2 RIPs from *Sambucus*, ebulin l and nigrin b, bind to cells to a lesser extent than ricin [4,31]. Furthermore, these RIPs have lower affinities for galactose than ricin [4,13,65]. The differential affinity of

RIPs from *Sambucus* for galactosides on cell surfaces may determine its intracellular fate and possibly its cytotoxicity. Thus, it has been reported that, unlike ricin, type 2 RIPs from *Sambucus* follow a weakly productive Golgi-independent pathway to the cytosol to intoxicate the cells [4,14,31]. An important observation when comparing the toxicity of nigrin l, nigrin-RPs, and nigritin l towards COLO 320 and HeLa cells is that nigrin l was the most active toxin. Nigrin l is a galactose binding lectin, whereas nigrin-RPs 1–3 failed to bind to the affinity matrix of AT-Sepharose and to agglutinate erythrocytes. Therefore, the low cytotoxicity of nigrin-RPs compared to nigrin l could be related to deficient sugar binding domains, which is the major difference with nigrin l. According to this, the toxicity of nigrin-RPs is comparable to that of the type 1 RIP nigritin l that lacks a B chain. Several studies underline that the cytotoxicity of RIPs is associated with their ability to induce apoptosis [28]. We also found that RIPs from *S. nigra* leaves induced apoptosis. HeLa cells treated with the RIPs showed apoptotic morphological features and, in COLO 320 cells, nigrin l treatment led to oligonucleosomal DNA fragmentation. One of the unresolved questions is whether rRNA damage leading to the inhibition of protein synthesis is solely responsible for this RIP-induced apoptosis. On the one hand, the ribosome may not be the only substrate of RIP action and, on the other hand, in the case of type 2 RIPs, apoptosis might also be induced by lectin binding to specific glycosylated proteins on the cell membrane, leading to activation of cell death factor receptors.

#### **4. Conclusions**

This work contributes to expanding our knowledge of the family of RIPs and RIPrelated lectins produced by *S. nigra*, with eight new proteins found in the leaves. Our results demonstrate the presence in this tissue of a type 2 RIP and two related lectins specific for galactose, four type 2 RIPs with deficient sugar binding domains, and one type 1 RIP.

Although initially only SNAI has been used, with the discovery and study of other proteins of the genus *Sambucus*, its use has been extended to other proteins, such as SSA, ebulin l, nigrin b, and SNAI'. The study of RIPs and related lectins of elderberry leaves completes the knowledge of this type of proteins in this species, and opens new perspectives, not only in the study of the biological functions attributed to them, but also in their use in biomedicine and agriculture.

New techniques such as protein structure prediction based on deep learning and neural networks, and peptide mapping based on high-resolution nanoLC–tandem mass spectrometry, can be very useful tools to further advance our knowledge of these types of proteins.

#### **5. Materials and Methods**

#### *5.1. Materials*

The sources of the chemicals were described previously [16]. Leaves from elder were harvested at Cobos de Cerrato (Palencia, Spain) in early summer. CM-Sepharose FF, Q-Sepharose FF, CM-Sepharose FF, Sepharose 6 B, and Superdex−75 HiLoad 26/60 columns were purchased from GE Healthcare (Barcelona, Spain). The acid-treated Sepharose (AT-Sepharose) was prepared as described in [66], treating the Sepharose 6 B with 0.1 N HCl at 50 ◦C for 3 h. The gel was then washed with water (Elix 5, Millipore) until a neutral pH was obtained, and stored in water at 4 ◦C until it was used. Tosyl phenylalanyl chloromethyl ketone (TPCK)-treated trypsin was purchased from Merk Life Science S.r.l. (Milan, Italy). Acetonitrile (CH3CN), formic acid (FA), and water (LC–MS grade) were from Fisher Scientific Italia (Milan, Italy). Century™-Plus RNA Markers were purchased from Fisher Scientific (Madrid, Spain). Rabbit reticulocyte lysate system (nuclease-treated) was purchased from Promega Biotech Iberica S.L. (Alcobendas, Madrid, Spain).

#### *5.2. Cell Lines and Culture*

COLO 320 (human colon carcinoma) and HeLa (human cervix epithelioid carcinoma) cells were obtained from the European Culture Collection (ECACC) and grown in RPMI

1640 medium (GIBCO BRL, Barcelona, Spain) supplemented with 10% fetal bovine serum (FBS), 100 U/mL penicillin, and 0.1 mg/mL streptomycin, under 5% CO2 at 37 ◦C.

#### *5.3. Methods*

#### 5.3.1. Preparation of Crude and Acidified Extracts and Chromatography Using SP-Sepharose

Two extracts were prepared from 700 and 540 g of *S. nigra* leaves. The leaves were crushed with dry ice in a crusher (Sammic Cutter K−52) and extracted with eight volumes of PBS (5 mM sodium phosphate, pH 7.5, 0.14 M NaCl) overnight at 4 ◦C. The crude extract was clarified by filtering it through a nylon mesh and then centrifuging it for 30 min at 9000 rpm in a JA−10 rotor (12,900× *g*) at 2 ◦C. The crude extract was acidified by adding glacial acetic acid to a pH of 4.05, and clarified again by filtering it through a nylon mesh and centrifuging it under the same conditions. The acidified crude extract was loaded onto a SP-Sepharose FF column (25 × 5 cm = 491 mL) equilibrated with 10 mM sodium acetate (pH 4.5). The column was washed at a flow rate of 7 mL/min with the same buffer until the absorbance at 280 nm of eluent dropped to almost zero. Proteins were first eluted with 5 mM sodium phosphate (pH 6.66) and then with the same buffer containing 1 M NaCl. Protein elution was controlled by measuring the absorbance at 280 nm, and the fractions eluted with sodium phosphate and NaCl were collected separately.

#### 5.3.2. Purification of Nigrin l, SNAlm, and SNAld

The protein eluted with sodium phosphate from the first preparation (1116 mg) was supplemented with 0.28 M NaCl and subjected to chromatography using AT-Sepharose (25 × 5 cm = 137 mL) equilibrated with 5 mM sodium phosphate (pH 7.5) containing 0.28 M NaCl. The column was kept at 0 ◦C and three identical chromatographies were performed, each loading 372 mg of protein. The column was washed at a flow rate of 4.5 mL/min with sodium phosphate 5 mM (pH 7.5) containing 0.28 M NaCl until the absorbance at 280 nm dropped to almost zero, and was eluted with the same buffer containing 0.2 M lactose. The eluate from the three chromatographies (218 mg of protein) was combined and concentrated up to 17 mL using an Amicon YM10 membrane. Next, three aliquots were prepared, each of which was subjected to molecular exclusion chromatography using a Superdex 75 HiLoad 26/60 (60 × 2.6 cm = 319 mL) column equilibrated with PBS at a flow rate of 2 mL/min. Fractions of 5 mL were collected, and their composition was determined by SDS-PAGE in the presence and absence of 2-mercaptoethanol. Fractions containing SNAlm were collected, dialyzed against water, frozen, and freeze dried; 84 mg of lyophilized protein was obtained. Fractions containing nigrin l and SNAld were collected, concentrated, and rechromatographed on the same column to remove traces of SNAlm, dialyzed. Subsequently, these fractions, in 5 mM sodium phosphate (pH 7.5), were subjected to anion exchange chromatography using Q-Sepharose FF (5 × 1.6 cm = 10 mL) equilibrated with 5 mM sodium phosphate (pH 7.5) at a flow rate of 4 mL/min. After loading the sample and washing with sodium phosphate buffer, the protein was eluted with a linear gradient of 600 mL of NaCl from 0 to 0.6 M. Fractions of 8 mL were collected, and the fractions containing nigrin l were pooled together, dialyzed against water, frozen, and freeze dried. The SNAld was further purified by repeating the same chromatography procedure; 14 mg of nigrin l and 10 mg of SNAld were obtained.

#### 5.3.3. Purification of Nigrin-RPs 1, 2, 3, and 4, and Nigritin l

The protein eluted with NaCl from the second preparation (1820 mg) was dialyzed and subjected to cation exchange chromatography using CM-Sepharose FF (7 × 2.6 cm = 37 mL) equilibrated with 5 mM sodium phosphate (pH 6.66) at a flow rate of 7 mL/min. After loading the sample and washing with sodium phosphate, the protein was eluted with a linear gradient of 1596 mL of NaCl from 0 to 0.3 M. Fractions of 10.5 mL were collected, which were tested via protein synthesis, and analyzed by SDS-PAGE. Fractions containing nigrin-RPs 1–4 and nigritin l were placed together, concentrated using an Amicon YM10

membrane, and separately subjected to molecular exclusion chromatographies through a Superdex 75 HiLoad 26/60 column equilibrated with PBS at a flow rate of 2 mL/min. Fractions of 5 mL were collected, and their composition was determined by SDS-PAGE in the presence and absence of 2-mercaptoethanol. Nigrin-RP1 was subjected to chromatography using AT-Sepharose (5 × 4 cm = 20 mL) to remove traces of nigrin l; the fraction not retained by the AT-Sepharose was dialyzed against water and freeze dried, obtaining 12 mg of lyophilized protein. Both the fractions containing nigrin-RP2 and those containing nigrin-RP4 were re-chromatographed using a Superdex 75 HiLoad column to eliminate traces contaminating the other protein, obtaining 14 and 0.15 mg of nigrin-RP2 and nigrin-RP4, respectively. The purification of nigrin-RP3 was conducted using the first preparation. After molecular exclusion chromatography, the protein was dialyzed and purified to homogeneity by chromatography using SP-Sepharose FF (4.5 × 1 cm = 3.5 mL) equilibrated with 5 mM sodium phosphate (pH 7.5) at a flow rate of 1 mL/min. After loading the sample and washing with sodium phosphate, the protein was eluted with a linear gradient of 140 mL of NaCl from 0 to 0.2 M. Fractions of 2 mL were collected and the fractions containing nigrin-RP3 were placed together, dialyzed against water, frozen, and freeze dried, obtaining 2 mg of lyophilized protein. Nigritin l was re-chromatographed using a Superdex 75 HiLoad column to remove traces of contaminants, dialyzed, and freeze dried, obtaining 8 mg of lyophilized protein.

#### 5.3.4. Analytical Procedures

Protein concentrations were determined using the spectrophotometric method of Kalb and Bernlohr [67]. Analyses of proteins by SDS-PAGE were carried out as described elsewhere [68] using 12% acrylamide gels and the Hoefer™ MiniVE system (Thermo Fisher Scientific-ES, Madrid, Spain). The glycosylation analysis was performed in gel after SDS-PAGE with a Mini-Protean II system (Bio-Rad; Milan, Italy) using the Pro-Q emerald 300 Glycoprot Probes Kombo (Life Technologies, Monza, Italy). Glycosylated proteins were visualized using the ChemiDoc™ XRS system.

#### 5.3.5. Assays of Cell-Free Protein Synthesis

The effect of RIPs on protein synthesis was determined through a coupled transcription– translation in vitro assay using a rabbit reticulocyte lysate system [13]. The reaction mixture contained 0.6 μL of rabbit reticulocyte lysate and 5.8 μL of a mixture of the following components: 4.6 U ribonuclease inhibitor, 2.3 U T7 RNA polymerase, 0.2 μg luciferase T7 plasmid, rNTPs (0.4 mM each), amino acids (2 μM each), 10 mM Tris–HCl (pH 7.8), 0.2 mM spermidine, 28 mM KCl, 1 mM MgCl2, and nuclease-free water. The mixtures were incubated at 30 ◦C for 10 min and placed on ice. Then, 1.6 μL of either water or different protein concentrations were added and the sample mixture was incubated at 30 ◦C for 40 min. Subsequently, 25 μL water was added and mixed with 28 μL of Luciferase Assay Reagent (Promega, Alcobendas, Madrid, Spain) at room temperature. Luminescence was determined with a Junior LB 9509 luminometer (Berthold Technologies GmbH & Co. KG, Bad Wildbad, Germany). Three experiments were conducted in duplicate, and IC50 (concentration that inhibits 50% protein synthesis) values were calculated by linear regression.

5.3.6. rRNA N-Glycosylase Assays on Rabbit Reticulocyte, Yeast, Bacterium Lysates, and HeLa Cells

rRNA N-glycosylase assays were conducted as described elsewhere [16]. Rabbit reticulocyte lysate (40 μL) was incubated with 5 μg of RIP at 30 ◦C for 1 h. N-glycosylase activity on *Saccharomyces cerevisiae* ribosomes was assayed in 50 μL samples of S−30 lysate from yeast in 10 mM Tris–HCl buffer (pH 7.6) containing 10 mM KCl, 10 mM magnesium acetate, and 6 mM 2-mercaptoethanol, which was incubated with 5 μg of RIPs at 30 ◦C for 1 h. N-glycosylase activity on *Micrococcus lysodeikticus* ribosomes was assayed using 100 μL of bacterial lysate samples in 20 mM Tris–HCl buffer (pH 7.8), which were incubated with 5 μg of RIP at 30 ◦C for 1 h. After treatment, the RNA was extracted by phenolization, treated with 1 M aniline acetate (pH 4.5), and precipitated with ethanol. HeLa cells (1 × 106/plate) were incubated for 48 h in the presence of 40 nM of nigrin l. After treatment, cells were harvested by centrifugation at 1000× *g* for 5 min. The pellets were lysed, and the RNA was isolated following the instruction of the RNeasy Mini Kit (Qiagen GmbH, Hilden, Germany). RNA was treated with 1 M aniline acetate (pH 4.5) for 10 min at 0 ◦C and precipitated with ethanol. The RNAs were subjected to electrophoresis at 15 mA for 2 h (rabbit and HeLa cells) or 1 h 30 min (yeast and bacterium) in a 7 M urea/5% (*w*/*v*) polyacrylamide gel and stained with GelRed nucleic acid stain (Biotium Inc., Hayward, CA, USA) [16].

#### 5.3.7. Adenine Polynucleotide Glycosylase Activity on Salmon Sperm DNA and Tobacco Mosaic Virus (TMV) RNA

The adenine release was measured according to the method reported elsewhere [69] with a few modifications. First, 10 μg of salmon sperm DNA was incubated with 5 μg of RIP in 300 μL of a reaction mixture containing 100 mM KCl and 50 mM magnesium acetate (pH 4), at 30 ◦C for 2 h. After incubation, the DNA was precipitated with ethanol at −80 ◦C for 3 h and centrifugated at 13,000 rpm for 15 min. Adenine released from RIP-treated DNA was determined in the supernatants spectrophotometrically at 260 nm. Analysis of the adenine polynucleotide glycosylase activity on tobacco mosaic virus (TMV) RNA was carried out as described elsewhere [18].

#### 5.3.8. DNA Cleavage Experiments

Nicking activity experiments were performed as previously reported [18]. Each reaction contained 5 μg of RIP and 200 ng of pCR2.1 DNA in a final volume of 10 μL, comprised of 10 mM Tris–HCl, 5 mM MgCl2, 50 mM NaCl, and 50 mM KCl, pH 7.8. Samples were incubated for 2 h at 30 ◦C, run on agarose gel (0.8%) in TAE buffer (0.04 M Tris, 0.04 M acetate, 1 mM EDTA, pH 8.0) and visualized by GelRed nucleic acid staining (Biotium Inc., Hayward, CA, USA).

#### 5.3.9. Tryptic Digestion and Sample Preparation for MS/MS Analyses

Aliquots of protein samples (50 μg) were reduced with 20 mM dithiothreitol (DTT, 5 min at 95 ◦C) and alkylated with 20 mM iodoacetamide (IAA, 30 min, in the dark, at room temperature). Enzymatic hydrolyses were performed on reduced and alkylated samples by adding TPCK-treated trypsin with an enzyme/substrate (E/S) ratio of 1:200 (*w*/*w*) for 3 h, 1:100 for 16 h, and 1:50 for 4 h at 37 ◦C.

#### 5.3.10. High-Resolution NanoLC–Tandem Mass Spectrometry

Mass spectrometry analyses on tryptic samples (500 fmol) were performed on a Q Exactive Orbitrap mass spectrometer equipped with an EASY-Spray nano-electrospray ion source (Thermo Fisher Scientific, Bremen, Germany) and coupled with a Thermo Scientific Dionex UltiMate 3000 RSLCnano system (Thermo Fisher Scientific). Solvent composition was 0.1% formic acid in water (solvent A) and 0.1% formic acid in acetonitrile (solvent B). Peptides were loaded on a trapping PepMap™100 μCartridge Column C18 (300 μm × 0.5 cm, 5 μm, 100 angstroms) and desalted with solvent A for 3 min at a flow rate of 10 μL/min. After trapping, eluted peptides were separated on an EASY-Spray analytical column (50 cm × 75 μm ID PepMap RSLC C18, 3 μm, 100 angstroms) and heated to 35 ◦C at a flow rate of 300 nL/min using the following gradient: 4% B for 3 min, from 4% to 55% B in 60 min, from 55% to 70% B in 10 min, and from 70% to 95% B in 2 min. Eluting peptides were analyzed on the Q-Exactive mass spectrometer operating in positive polarity mode with capillary temperature of 280 ◦C and a potential of 1.9 kV applied to the capillary probe. Full MS survey scan resolution was set to 70,000 with an automatic gain control (AGC) target value of 3 × 106 for a scan range of 375–1500 *m/z* and maximum ion injection time (IT) of 100 ms. The mass (*m*/*z*) 445.12003 was used as lock mass. A data-dependent top 5 method was operated, during which higher-energy collisional dissociation (HCD) spectra were obtained at 17,500 MS2 resolution with AGC target of 1 × <sup>10</sup><sup>5</sup> for a scan range of 200–2000 *m*/*z*, maximum IT of 55 ms, 2 *m*/*z* isolation width, and normalized collisional energy (NCE) of 27. Precursor ions targeted for HCD were dynamically excluded for 15 s. Full scans and Orbitrap MS/MS scans were acquired in profile mode, whereas ion trap mass spectra were acquired in centroid mode. Charge state recognition was enabled by excluding unassigned charge states.

#### 5.3.11. Data Processing

The acquired raw files were analyzed with Proteome Discoverer 2.4 software (Thermo Fisher Scientific, Rockford, IL, USA) using the SEQUEST HT search engine. The HCD MS/MS spectra were searched against the whole UniProt\_SwissProt KB database and against a homemade database including *S. nigra* RIP sequences deposited into the NCBI\_ GeneBank\_NIH assuming trypsin (Full) as digestion enzyme with two allowed numbers of missed cleavage sites. The mass tolerances were set to 10 ppm and 0.02 Da for precursor and fragment ions, respectively. Oxidation of methionine (+15.995 Da) was set as dynamic modification, and carbamidomethylation of cysteine (+57.021 Da) as static modification. False discovery rates (FDRs) for peptide spectral matches (PSMs) were calculated and filtered using the Target Decoy PSM Validator Node in Proteome Discoverer. The Target Decoy PSM Validator Node specifies the PSM confidences based on dynamic score-based thresholds. It calculates the node-dependent score thresholds needed to determine the FDRs, which are provided as input parameters of the node. The Target Decoy PSM Validator was run with the following settings: maximum Delta Cn of 0.05, a strict target FDR of 0.01, a relaxed target FDR of 0.05, and validation based on q-value. The Protein FDR Validator Node in Proteome Discoverer was used to classify protein identifications based on q-value. Proteins with a q-value of <0.01 were classified as high-confidence identifications and proteins with a q-value of 0.01–0.05 were classified as medium-confidence identifications. Only proteins identified with medium or high confidence were retained, resulting in an overall FDR of 5%. All MS/MS spectra were manually validated to assign the best PSM to peptide sequences. When multiple PSM were mapped on the same peptide sequence, those with the highest accuracy were selected.

#### 5.3.12. Hemagglutination Activity and Carbohydrate Binding Properties

Hemagglutination activity (HA) was assayed as described elsewhere [25]. The HA was determined in microtiter plates. Each well contained 50 μL of serial dilutions of the proteins and 50 μL of 1% erythrocyte suspension, and the plates were incubated for 1 h at room temperature. The minimum concentration of protein causing complete agglutination was visually evaluated. For hemagglutination inhibition assay, six sugars (D-glucose, Dgalactose, D-fructose, D-mannose, L-fucose, and lactose) were tested for their ability to inhibit the HA of the proteins. Each well contained 50 μL of carbohydrates serially diluted, and 25 μL of the proteins at a concentration one titer higher than the HA dose. An equal volume of 2% erythrocyte suspension (25 μL) was added to each well and incubated for 1 h at room temperature. The minimum concentration of the tested sugars that completely inhibited HA activity was determined.

#### 5.3.13. Protein Structure Studies and Graphical Representation

The structure of ricin (accession number 2 AAI) is available in the Protein Data Bank (https://www.rcsb.org/ (accessed on 26 April 2022). The three-dimensional structural modeling of nigrin l, SNAlm, and SNAld was carried out with the AlphaFold2 software following the instructions of the website https://colab.research.google.com/ github/sokrypton/ColabFold/blob/main/AlphaFold2.ipynb#scrollTo=G4yBrceuFbf3 (accessed on 26 May 2022) [23]. The study representations and graphs of protein structures were constructed with the help of the Discovery Studio Visualizer suite (v21.1.0) (https://www.3dsbiovia.com/ (accessed on 26 April 2022). The SNAld dimer model was

carried out on the SymmDock server (http://bioinfo3d.cs.tau.ac.il/SymmDock/ (accessed on 19 June 2022)) [70].

#### 5.3.14. Molecular Docking

The structures of β-D-galactose (PubChem CID 439353) and β-lactose (PubChem CID 6134) are available in the PubChem database (https://pubchem.ncbi.nlm.nih.gov/ (accessed on 10 June 2022)) [71]. Docking was carried out using Autodock 4.2 (http: //autodock.scripps.edu/ (accessed on 15 October 2021)) [72], as described elsewhere [25]. The docking of D-galactose was performed on a grid of 120 × 120 × 120 points, with the addition of a central grid point. The grid was centered on the centroid of the pyranosic ring of galactose at the 1α or 2γ sites of the ricin structure (accession number 2AAI). The grid spacing was 0.125 angstroms, which led to a grid of 15 × 15 × 15 angstroms. For each molecule, 100 docking runs were performed. The 100 docking poses generated were clustered by root mean square (RMS) difference with a cutoff value of 0.5 angstroms for each case. The top-ranked pose of the most populated clusters was retained and further analyzed with the Discovery Studio Visualizer suite (v21.1.0). β-lactose docking was performed as indicated for D-galactose, but using a grid of 124 × 124 × 124 points and a grid spacing of 0.180 angstroms, resulting in a grid of 22.32 × 22.32 × 22.32 angstroms. The grid was centered on the centroid of the pyranosic rings of lactose at the 1α or 2γ sites of the ricin structure. The 100 docking poses generated were compared with those obtained for D-galactose, and a pose was chosen that (coinciding with the best solution of docking with D-galactose) had a lower binding energy.

#### 5.3.15. Cell Viability and DNA Fragmentation Analyses

Cell viability was determined using a colorimetric assay based on the cleavage of the tetrazolium salt WST-1 to formazan by mitochondrial dehydrogenases in viable cells. First, <sup>3</sup> × <sup>10</sup><sup>3</sup> COLO 320 or HeLa cells in 0.1 mL of medium were seeded in 96-well plates and incubated at 37 ◦C under 5% CO2 in the absence or the presence of RIP for 48 h. Next, the cells were incubated for another 2 h with 10 μL/well of the cell proliferation reagent WST-1 at 37 ◦C under 5% CO2. The absorbance of the samples was measured using a microtiter plate reader set at 450 nm with 620 nm as reference (ELISA reader Multiskan). The absorbance of cultures without cells was subtracted as background. For the DNA fragmentation analysis, COLO 320 cells (1 × 106/plate) were incubated for 72 h in the presence of 40 nM nigrin l. After treatment, cells were harvested by centrifugation (1000× *g* for 5 min). The pellets were lysed in 50 mM Tris–HCl, pH 8.0, containing 10 mM EDTA and 0.5% SDS, and the DNA was isolated following the instruction of the Genomic Prep Cells and Tissue DNA Isolation Kit (GE Healthcare, Barcelona, Spain). DNA electrophoresis was carried out in 1.8% agarose gels using TBE buffer (0.089 M Tris, 0.089 M boric acid, 2 mM EDTA, pH 8.0), and 4.0 μg of DNA was electrophoresed and stained with GelRed (Biotium Inc., Hayward, CA, USA) and visualized with an ultraviolet lamp.

**Supplementary Materials:** The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/toxins14090611/s1. Table S1: Amino acid sequences of tryptic peptides from SNAlm obtained by high-resolution nanoLC–tandem mass spectrometry and mapped on SNAlm (AC: AAN86132). Table S2: Amino acid sequences of tryptic peptides from SNAld obtained by high-resolution nanoLC–tandem mass spectrometry and mapped on SNAld (AC: AAN86131). Table S3: Amino acid sequences of tryptic peptides from nigrin l obtained by high-resolution nanoLC– tandem mass spectrometry and mapped on nigrin l (AC: AAN86130). Table S4: Amino acid sequences of tryptic peptides from nigrin-RP1 obtained by high-resolution nanoLC–tandem mass spectrometry and mapped on SNLRP2 (AC: AAC49672). Table S5: Amino acid sequences of tryptic peptides from nigrin-RP2 obtained by high-resolution nanoLC–tandem mass spectrometry and mapped on SNLRP2 (AC: AAC49672). Table S6: Amino acid sequences of tryptic peptides from nigrin-RP3 obtained by high-resolution nanoLC–tandem mass spectrometry and mapped on SNLRP1 (AC: AAC49673). Table S7: Inhibition of the hemagglutination activity of nigrin l, SNAlm, and SNAld by sugars compared with the estimated free energy of binding to the sugar binding sites. Figure S1: Effect of nigrin l, nigrin-RPs 1–4, and nigritin l on protein synthesis. Figure S2: Representative MS/MS fragmentation spectra of SNAlm peptides mapped on the protein SNAlm (AC: AAN86132). Figure S3: Representative MS/MS fragmentation spectra of SNAld peptides mapped on the protein SNAld (AC: AAN86131). Figure S4: Representative MS/MS fragmentation spectra of nigrin l peptides. Figure S5: Representative MS/MS fragmentation spectra of nigrin-RP1 peptides. Figure S6: Representative MS/MS fragmentation spectra of nigrin-RP2 peptides. Figure S7: Representative MS/MS fragmentation spectra of nigrin-RP3 peptides. Figure S8: Comparison of the binding of D-galactose and lactose to the 1α site of nigrin l, SNAlm, and SNAld. Figure S9: Comparison of the binding of D-galactose and lactose to the 2γ site of nigrin l, SNAlm, and SNAld. Figure S10: Structure of the SNAld dimer with D-galactose bound to 1α sites.

**Author Contributions:** Conceptualization, J.M.F. and L.C.; methodology, R.I., R.R., J.M.F. and L.C.; validation, A.C., J.M.F. and L.C.; formal analysis, R.I., R.R., A.C. and L.C.; investigation, R.I., R.R., N.L., M.V., A.C., A.D.M., J.M.F. and L.C.; resources, R.I., A.C., A.D.M., J.M.F. and L.C.; data curation, R.I., R.R., A.B., J.M.F. and L.C.; writing—original draft preparation, J.M.F. and L.C.; writing—review and editing, R.I., R.R., N.L., A.C., A.B., J.M.F. and L.C.; visualization, J.M.F. and L.C.; supervision, R.I., A.C., A.D.M., A.B., J.M.F. and L.C.; project administration, A.C., A.D.M., A.B., J.M.F. and L.C.; funding acquisition, A.C., A.B., J.M.F. and L.C. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was funded by grants BIO39/VA39/14 and BIO/VA17/15 (Consejería de Sanidad, Junta de Castilla y León) to L.C.; grant VA033G19 (Consejería de Educación, Junta de Castilla y León) to the GIR ProtIBio; MISE, project NUTRABEST PON I&C 2014–2020 Prog. n. F/200050/01–03/X45; and the Pallotti Legacies for Cancer Research.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Data are available upon request; please contact the contributing authors.

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

#### **References**


## *Article* **Characterization of Lung Injury following Abrin Pulmonary Intoxication in Mice: Comparison to Ricin Poisoning**

**Anita Sapoznikov \*, Yoav Gal, Ron Alcalay, Yentl Evgy, Tamar Sabo, Chanoch Kronman and Reut Falach \***

Department of Biochemistry and Molecular Genetics, Israel Institute for Biological Research, Ness-Ziona 74100, Israel

**\*** Correspondence: anitas@iibr.gov.il (A.S.); reutf@iibr.gov.il (R.F.); Tel.: +972-89381847 (A.S.); +972-89381522 (R.F.)

**Abstract:** Abrin is a highly toxic protein obtained from the seeds of the rosary pea plant *Abrus precatorius*, and it is closely related to ricin in terms of its structure and chemical properties. Both toxins inhibit ribosomal function, halt protein synthesis and lead to cellular death. The major clinical manifestations following pulmonary exposure to these toxins consist of severe lung inflammation and consequent respiratory insufficiency. Despite the high similarity between abrin and ricin in terms of disease progression, the ability to protect mice against these toxins by postexposure antibodymediated treatment differs significantly, with a markedly higher level of protection achieved against abrin intoxication. In this study, we conducted an in-depth comparison between the kinetics of in vivo abrin and ricin intoxication in a murine model. The data demonstrated differential binding of abrin and ricin to the parenchymal cells of the lungs. Accordingly, toxin-mediated injury to the nonhematopoietic compartment was shown to be markedly lower in the case of abrin intoxication. Thus, profiling of alveolar epithelial cells demonstrated that although toxin-induced damage was restricted to alveolar epithelial type II cells following abrin intoxication, as previously reported for ricin, it was less pronounced. Furthermore, unlike following ricin intoxication, no direct damage was detected in the lung endothelial cell population following abrin exposure. Reduced impairment of intercellular junction molecules following abrin intoxication was detected as well. In contrast, similar damage to the endothelial surface glycocalyx layer was observed for the two toxins. We assume that the reduced damage to the lung stroma, which maintains a higher level of tissue integrity following pulmonary exposure to abrin compared to ricin, contributes to the high efficiency of the anti-abrin antibody treatment at late time points after exposure.

**Keywords:** abrin; ricin; intranasal; lungs; alveolar epithelial type II cells; neutrophils; alveolar–capillary barrier; junction proteins; glycocalyx

**Key Contribution:** Pulmonary exposure of mice to a lethal dose of abrin induces less pronounced damage to the lung stroma and a reduced impairment of intercellular junction molecules in comparison to ricin.

#### **1. Introduction**

The family of ribosome-inactivating proteins (RIPs) groups all enzymes (EC.3.2.2.22) with a so-called RIP domain which comprises N-glycosylase activity and enables these proteins to catalytically inactivate ribosomes. The highest number of plant RIPs has been found in angiosperm plants [1–4], including fungi [3,5,6], algae [7] and bacteria [8]. Structurally, plant RIPs can be divided into two main groups, depending on the presence or absence of a quaternary structure. Type 1 RIPs (~30 kDa) are single-chain proteins with enzymatic action, whereas type 2 RIPs (~60 kDa) consist of an enzymatically active A-chain linked to a B-chain with lectinic properties through a disulfide bridge. The B-chain binds to carbohydrates on the cell surface, allowing A-chain cell internalization. The absence of a lectinic chain prevents type 1 RIPs from binding to the cell, which are consequently less

**Citation:** Sapoznikov, A.; Gal, Y.; Alcalay, R.; Evgy, Y.; Sabo, T.; Kronman, C.; Falach, R. Characterization of Lung Injury following Abrin Pulmonary Intoxication in Mice: Comparison to Ricin Poisoning. *Toxins* **2022**, *14*, 614. https://doi.org/10.3390/ toxins14090614

Received: 10 August 2022 Accepted: 31 August 2022 Published: 2 September 2022

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

**Copyright:** © 2022 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 (https:// creativecommons.org/licenses/by/ 4.0/).

toxic with respect to type 2 RIPs, due to the difficulty of entering the cell [3]. Moreover, a third group of RIPs, known as type 3 RIPs, consists of a type-1-like N-terminal domain with N-glycosylase activity, covalently linked to a C-terminal domain with an unknown function [9].

The toxins abrin and ricin produced from the seeds of *Abrus precatorius* and *Ricinus communis*, respectively, are classified as type II RIPs and consist of an enzymatically active A-chain disulfide linked to a B-chain. The B-chain is a galactose-specific lectin that is responsible for the binding of toxins to glycoproteins or glycolipids on the surface of cells to promote endocytosis of the toxin [10]. Receptor-dependent internalization of the toxins involves retrograde transport to the endoplasmic reticulum, where the disulfide bond connecting the A and B subunits is reduced [11], allowing the release of the catalytically active A-chain into the cytoplasm [12]. The A subunit of both toxins is an RNA N-glycosylase that catalyzes the site-specific release of an essential adenine moiety, located in a highly conserved stem–loop within the small subunit of the ribosome 28S rRNA [13–15]. The irreversible depurination of the stem–loop by the A subunit prevents the binding of elongation factors to ribosomes, thereby inhibiting protein synthesis and eventually causing cell death [16,17].

Abrin's potential use as a chemical weapon stems from its high toxicity, together with the fact that it can be isolated from jequirity beans at a low cost by a relatively simple procedure. The relatively low-scale cultivation of jequirity plants compared with *Ricinus* (castor oil) plants would suggest a smaller and more focused terrorist-type chemical attack with abrin. One possible scenario is that once isolated, abrin could be aerosolized as a dry powder. Studies have found that the overall pattern and time course of damage following inhalation were similar for ricin and abrin and characterized by rapidly progressive and overwhelming pulmonary edema accompanied by acute destructive alveolitis and necrosis/apoptosis of the lower respiratory tract epithelium accounting for the majority of deaths [18–21]. The clinical manifestations following pulmonary (intranasal) exposure to these toxins were entirely restricted to the lungs and characterized by severe pulmonary edematous inflammation, neutrophil recruitment and development of a proinflammatory cytokine storm [22,23]. Pulmonary (intranasal) ricin and abrin intoxication in mice are similar with regard to pathological features and kinetics. However, despite their resemblance, the ability to protect mice against ricin and abrin intoxication by postexposure antibodymediated treatment differs drastically. Rabbit-derived polyclonal anti-ricin antibody-based treatment showed almost complete protection a few hours after exposure to a lethal ricin dose; however, when this treatment was delayed to 24 h after intoxication, only one-third of the mice survived [22]. In contrast, the intranasal administration of polyclonal anti-abrin antibodies to mice even as late as 72 h postexposure to a lethal dose of abrin conferred exceedingly high-level protection [23]. Interestingly, the efficient protection by polyclonal anti-abrin antibodies cannot be attributed to the specific neutralization of a particular A or B subunit of the toxin, as antibodies raised against chimeric toxins of either an AabrinBricin or AricinBabrin structure conferred exceptionally high protection levels to mice following intranasal exposure to a lethal dose of abrin [24].

In view of these findings, we characterized and quantified the cellular and molecular changes in murine lung tissue following pulmonary exposure to abrin, as compared to ricin intoxication, to delineate toxin-specific patho-physiologic factors that may play a role in determining the differential ability to protect against abrin and ricin by postexposure antibody administration.

#### **2. Results**

#### *2.1. Differential Binding of Abrin to Lung Cell Populations*

Previously, we showed that following intranasal intoxication of mice with a lethal dose of ricin, the toxin binds to alveolar macrophages (AMs) and dendritic cells (DCs) of the hematopoietic compartment, as well as to the lung parenchyma, epithelial and endothelial cells [25]. We examined, in a similar system, the interactions between abrin and lung cells after intranasal intoxication of mice with fluorescently labeled abrin (abrin Alexa Fluor 488 (abrin AF488)) at a lethal dose of 2LD50. Abrin-associated cells could be visualized by flow cytometry 3 h following exposure to the toxin (Figure 1A). To determine the kinetics of toxin binding to individual cell populations in the lung, mice were intoxicated with abrin or ricin, and lung cells isolated at different time points thereafter were analyzed for toxin binding. In the hematopoietic compartment (CD45+ cells), peak binding of both toxins was detected as early as 3 h after intoxication, and at later time points, fewer cells were detected in association with the toxins. The kinetics of CD45<sup>+</sup> cell binding and the proportion of toxin-bound cells at all examined time points was similar for both toxins (Figure 1B). Among cells of hematopoietic origin, abrin exhibited prompt binding to AMs and DCs, with peak binding at 3 h after intoxication (Figure 1C). To analyze the correlation between the binding of abrin and its ability to eliminate cells by inhibiting protein synthesis, we quantified the number of AMs and DCs at different time points following intoxication. In contrast to ricin pulmonary intoxication, where AMs were significantly reduced 3 h after exposure [25], AMs were heavily reduced 6 h after abrin intoxication, and their numbers stayed low at later time points (24–72 h postexposure, Figure 1D). At 24 h after either ricin or abrin intoxication, the population of AMs comprised only ~40–50% of the initial population of AMs observed in naïve mice (Figure 1E). In a similar manner, the DC population was reduced starting from 6 h post abrin exposure (Figure 1F), in contrast to the significant reduction in these cells already at 3 h after ricin intoxication [25]. One day after intoxication with either of the toxins, the DC population in the lung consisted of only ~60–70% of the initial population measured in non-intoxicated mice (Figure 1G).

**Figure 1.** Kinetics of abrin binding to hematopoietic cells and alteration in cell populations in the lung. Mice were intranasally exposed to fluorescent abrin or ricin AF488 (20 or 14 μg/kg body weight, respectively),

and lung cells were isolated at 3, 6, 12, 18, 24, 48 and 72 h after exposure and analyzed by flow cytometry for toxin binding by detection of AF488<sup>+</sup> cells and different cell population counts in the lungs. (**A**) Dot plots represent abrin AF488 staining in lung cells isolated 3 h after abrin intoxication or in cells isolated from control mice. (**B**) Quantification of toxin-bound CD45+ cells at different time points following intoxication. A comparison between ricin and abrin intoxications (*n* = 3–9 mice in each group). (**C**) Quantification of abrin-bound AMs and DCs at different time points following abrin intoxication (*n* = 3–6 mice in each group). Quantification of AM (**D**) and DC (**F**) population sizes at different time points following abrin intoxication (10 μg/kg, *n* = 3–15 mice in each group; each point indicates individual mice). The results are depicted as the means ± SEMs. \*\* *p* < 0.01, \*\*\* *p* < 0.001 in comparison to nonintoxicated mice; n.s., not significant. Comparison between abrin and ricin [25] AMs (**E**) and DCs (**G**) (% of control) at 24 h postexposure to toxins.

Next, we analyzed the binding of both toxins to the parenchymal cell populations of the lung (CD45− cells). Interestingly, we found that binding to CD45− cells was considerably more pronounced in the case of ricin exposure. Thus, at 3 h after intoxication, twice as many CD45− cells were bound to ricin than to abrin, while at 6 h postexposure, almost 6-fold more cells were associated with ricin than with abrin. Higher levels of toxin binding to lung parenchymal cells following ricin exposure were further observed at all later time points (12–72 h postexposure, Figure 2A). Within the parenchymal compartment, we distinguished between vascular endothelial cells and alveolar epithelial cells. Surprisingly, examination of the endothelial cell population following abrin intoxication did not result in observable damage to the cells, as evidenced by the preserved cell numbers at all tested time points postexposure (Figure 2B). This was in contrast to ricin intoxication, where a ~25% reduction in endothelial cells was detected 48 h postexposure (Figure 2C). Profiling of epithelial cells demonstrated a significant reduction in these cells from 48 h to 72 h after abrin intoxication (Figure 2D). A comparison of ricin and abrin at the same time point showed that the damage to epithelial cells after ricin intoxication was more pronounced, displaying a loss of ~60% of the population, while a reduction in these cells following abrin pulmonary intoxication was no more than ~40% (Figure 2E). Examination of subsets of the epithelial cells demonstrated that although no change was found in alveolar epithelial type I cells after pulmonary abrin intoxication (Figure 2F), the number of alveolar epithelial type II cells decreased significantly by 48 h postexposure (Figure 2G). The reduction in the alveolar epithelial type II population was also manifested with immunohistochemistry by specific labeling of pro-surfactant C (pro-SPC). Immunostaining of the lungs 48 h after abrin intoxication for alveolar epithelial type I cells (anti-T1a) and endothelial cells (anti-CD31) further confirmed no injury to these populations, as comparable staining levels were detected both in nonintoxicated (control) and intoxicated lungs (Figure 2H).

#### *2.2. Pulmonary Exposure to Abrin and Ricin Induces Comparable Neutrophil Influx to the Lungs Accompanied by Lung Hyperpermeability*

One prominent hallmark of ricin-mediated pulmonary intoxication is the rapid and massive influx of neutrophils to the lungs, where they contribute to the developing inflammation yet may also cause tissue damage, thereby promoting ricin-mediated morbidity [26–28]. These neutrophils are refractive to ricin binding [25]. Examination of the binding of abrin to neutrophils in the lungs following intranasal exposure of mice demonstrated that this cell type did not bind abrin at any time point (3–72 h) tested after intoxication (Figure 3A,B). There were no abrin AF488-positive neutrophils in the lungs, which could be appreciated by the low mean fluorescent intensity (MFI) of fluorescent abrin in these cells at all time points after intoxication (Figure 3C). These results are in sharp contrast to AMs, which readily bound abrin, as shown by the increased abrin AF488 intensity (Figure 3B). Next, we monitored neutrophil numbers at different time points after abrin intoxication. Neutrophil counts that comprised ~2 × 106 cells in healthy mice were raised to 26 ± 4 × 106 cells and 50 ± 11 × 106 cells at 24 h and 72 h, respectively, after exposure to abrin (Figure 3D). Alignment between the elevation of neutrophils after ricin and abrin intoxications showed equal recruitment of these cells to the lungs 24–72 h postexposure

(Figure 3E). Since uncontrolled massive recruitment of neutrophils may cause tissue damage and promote permeability and edema, we measured lung permeability by the Evans blue dye (EBD) extravasation assay. To this end, mice were intravenously injected with EBD at different time points after intranasal exposure to abrin, lungs were harvested, and EBD was extracted and quantified. Pulmonary EBD levels were found to be elevated significantly at 48 and 72 h post abrin exposure (Figure 3F). These results indicate that similar to ricin pulmonary intoxication, exposure to abrin promotes comparable neutrophil influx into the lungs, which is accompanied by lung hyperpermeability.

**Figure 2.** Kinetics of abrin binding to parenchymal cells and alteration in cell populations in the lung. Mice were intranasally exposed to fluorescent abrin or ricin AF488 (20 or 14 μg/kg body weight, respectively), and lung cells were isolated at 24, 48 and 72 h and analyzed by flow cytometry for toxin

binding by detection of AF488+ cells and for different cell population counts in the lungs. (**A**) Quantification of toxin-bound CD45− cells at different time points following intoxication. Comparison between ricin and abrin intoxications (*n* = 3 mice per group). (**B**) Mice were intranasally exposed to abrin (10 μg/kg), lungs were removed at indicated time points, and endothelial and epithelial (**D**) numbers were determined by flow cytometry (*n* = 6–9 mice in each group). Comparison between the percent of endothelial (**C**) or epithelial (**E**) cells at 48 h post abrin exposure to the percent of these cells at the same time point after ricin intoxication [25]. (**F**) Quantification of alveolar epithelial type I (ATI) and alveolar epithelial type II (ATII) (**G**) cell populations at 48 h post abrin exposure (*n* = 8–17 mice in each group). (**H**) Immunofluorescence analysis of ATI (T1α, red) and ATII (pro-SPC, green) or endothelial cell (CD31, green) staining of lung tissue in nonintoxicated mice (control) versus abrin 48 h postexposed mice (blue, 4- ,6-diamidino-2-phenylidole (DAPI) staining of nuclei). Scale bar: 50 μm. The results are depicted as the means ± SEMs. \* *p* < 0.05, \*\* *p* < 0.01, \*\*\* *p* < 0.001; n.s., not significant. (**A**,**C**,**E**) Comparison between intoxications at each time point; (**B**,**D**,**F**,**G**) comparison with nonintoxicated mice.

**Figure 3.** Effect of exposure to abrin on neutrophils and lung permeability. Mice were intranasally exposed to fluorescent abrin AF488 (20 μg/kg body weight), and lung cells were isolated at 3, 6, 24, 48 and 72 h and analyzed for neutrophils by flow cytometry. (**A**) Dot plots represent abrin AF488 staining in neutrophils isolated 3 h after abrin intoxication or in cells isolated from control mice. (**B**) Abrin AF488 binding to neutrophils (black histograms) and AMs (red histograms) at different time points following abrin intoxication. (AMs are autofluorescent in the lungs and exhibit high background in control mice.) (**C**) MFI of abrin AF488 binding to neutrophils (*n* = 3–7 mice in each group). (**D**) Neutrophil count in the lungs at different time points following abrin intoxication. (**E**) Comparison between the increase in neutrophils after abrin and ricin [25] intoxication (*n* = 3–15 mice in each group). (**F**) Lung EBD extravasation following abrin intoxication. Control or abrin-intoxicated mice were intravenously injected with 50 mg/kg EBD at the indicated time points, and lungs were monitored for EBD content (*n* = 4–10 mice in each group). The results are depicted as the means ± SEMs. \*\*\* *p* < 0.001; n.s., not significant. In (**D**,**F**), comparison with nonintoxicated mice.

#### *2.3. Pulmonary Exposure to Abrin Leads to Inferior Impairment of Junction Proteins in the Lungs in Comparison to Ricin*

The integrity of the alveolar wall barrier depends on the intercellular junctions of the alveolar epithelial and capillary endothelial cells. Junction protein complexes formed by tight junctions (TJs), adherens junctions (AJs) and gap junctions (GJs) stabilize the connections between contiguous cells. Disruption of the integrity of these complexes results in increased permeability and the formation of lung edema [29–31]. Indeed, we have previously shown that in the case of ricin pulmonary intoxication, disruption of differential intercellular junction proteins leads to impairment of the alveolar–capillary barrier and to the development of lung edema, which in turn results in the impairment of oxygenation [27,32]. Occludin, a TJ protein, is expressed in the alveoli, bronchial epithelial cells and endothelial cells of blood vessels in healthy lungs. We did not detect any change in occludin expression at 24 and 48 h after abrin intoxication (Figure 4A,B), which was in contrast to ricin intoxication which triggered a two-fold reduction in occludin in the lungs (Figure 4C). Next, we examined the expression of VE-cadherin, which is essential for endothelial barrier integrity. The VE-cadherin level was significantly decreased following abrin exposure (Figure 4D,E), although in comparison to ricin intoxication, which brought nearly complete elimination of VE-cadherin at 3–6 h postexposure, the damage to this protein by abrin was only moderate, even at 24 h post abrin exposure (Figure 4F). Similar to VE-cadherin, claudin 18, which is expressed by alveolar epithelial cells, was also diminished after abrin intoxication (Figure 4G,H), but once again, its reduction was less pronounced than that recorded after pulmonary exposure to ricin (Figure 4I).

**Figure 4.** Alterations in occludin, VE-cadherin and claudin 18 in the lungs of abrin-intoxicated mice. Lungs of abrin-intoxicated (10 μg/kg body weight) mice were harvested at the indicated time points, and junction proteins were quantified by immunohistochemical analysis of lung sections. (**A**,**D**,**G**) Confocal microscopy scans of lung sections stained for occludin, VE-cadherin and claudin 18 (red), respectively, and identification of nuclei by DAPI (blue). (**B**,**E**,**H**) Scatterplots represent the immunofluorescence staining intensities of occludin, VE-cadherin and claudin 18, expressed as MFI (*n* = 3–5 mice in each group). Scale bar: 50 μm. (**C**,**F**,**I**) Comparison between the abrin- and ricininduced reduction in occludin (at 48 h), VE-cadherin and claudin 18 [27]. The results are depicted as the means ± SEMs. \* *p* < 0.05, \*\* *p* < 0.01, \*\*\* *p* < 0.001; n.s., not significant. In (**B**,**E**,**H**), comparison with nonintoxicated mice.

Examination of connexin 43, a GJ protein, showed that its level was heavily decreased at 6 and 24 h post abrin intoxication (Figure 5A,B). This pronounced decrease in connexin 43 correlated with its elimination after ricin intoxication (Figure 5C). Finally, we examined the expression of claudin 5, which is expressed in both endothelial and airway epithelial cells. The level of claudin 5 was significantly diminished 3 to 24 h after abrin intoxication (Figure 5D,E). As in the connexin 43 case, the reduction in claudin 5 following abrin intoxication was comparable to the level of the protein at the corresponding time points after ricin intoxication, as demonstrated by MFI (Figure 5F).

**Figure 5.** Alterations in connexin 43 and claudin 5 in the lungs of abrin-intoxicated mice. Lungs of abrin-intoxicated (10 μg/kg body weight) mice were harvested at the indicated time points, and junction proteins were quantified by immunohistochemical analysis of lung sections. (**A**,**D**) Confocal microscopy scans of lung sections stained for connexin 43 and claudin 5 (red) and identification of nuclei by DAPI (blue). (**B**,**E**) Scatterplots represent the immunofluorescence staining intensities of connexin 43 and claudin 5, expressed as MFI (*n* = 3–4 mice in each group). Scale bar: 50 μm. (**C**,**F**) Comparison between the abrin- and ricin-induced reduction in connexin 43 and claudin 5 [27]. The results are depicted as the means ± SEMs. \*\*\* *p* < 0.001. In (**B**,**E**), comparison with nonintoxicated mice; n.s., not significant.

#### *2.4. Pulmonary Exposure to Abrin- and Ricin-Induced Comparable Damage to the Endothelial Glycocalyx*

The endothelial glycocalyx is a complex layer of glycoproteins, proteoglycans and glycosaminoglycans that coat the luminal surface of the vascular endothelium. Hydrated glycosaminoglycans form a thick and rigid endothelial surface layer (ESL) that plays a key role in limiting vascular permeability and regulating leukocyte adhesion [33,34]. Because shedding of the ESL results in hyperpermeability and inappropriate leukocyte adhesion [35], we decided to evaluate the integrity of the ESL after pulmonary exposure to abrin in comparison to ricin intoxication. Mice were intranasally exposed to abrin or ricin, and soluble shed compounds of glycocalyx were analyzed in bronchoalveolar fluid (BALF) harvested at different time points. The detection of soluble glycocalyx compounds, which are indicative of endothelial glycocalyx degradation, negatively correlates with ESL thickness and positively correlates with vascular permeability [30,36]. Syndecan-1, heparan sulfate and hyaluronic acid are the main components whose shedding has been claimed to represent the endothelial glycocalyx state of health. An analysis of the hyaluronic acid levels, a ubiquitous glycosaminoglycan of the ECL, revealed elevated levels of this compound in the BAL of ricin- and abrin-intoxicated mice at 24 h postexposure, which

continued to increase at later time points, 48–72 h postexposure. No difference was found in the intensity of the shedding of hyaluronic acid between abrin- and ricin-intoxicated mice (Figure 6A). Monitoring of heparan sulfate, the predominant glycosaminoglycan, demonstrated a marked release of this component at 72 h post abrin and ricin exposures. As in the case of hyaluronic acid, heparan sulfate shedding was similar in response to both toxins at all indicated time points (Figure 6B). Despite the degradation of hyaluronic acid and heparan sulfate, we did not detect shedding of the transmembrane core protein of the glycocalyx, syndecan-1, following either abrin or ricin pulmonary exposure (Figure 6C).

**Figure 6.** Degradation of the glycocalyx in abrin- and ricin-exposed mice. Levels of soluble hyaluronic acid (**A**), heparan sulfate (**B**) and syndecan-1 (**C**) were determined in the BALF collected from abrinand ricin (10 or 7 μg/kg body weight, respectively)-exposed mice at the indicated time points (*n* = 4–9 mice in each group). The results are depicted as the means ± SEMs. \* *p* < 0.05, \*\* *p* < 0.01, \*\*\* *p* < 0.001; n.s., not significant. The comparison of each column with nonintoxicated mice and between abrin and ricin at each time point.

#### **3. Discussion**

The toxicity of abrin and ricin depends on the route of exposure, with inhalatory exposures considered the most fatal [16]. The clinical manifestation following intranasal exposure of mice to these toxins is the onset of localized yet severe pulmonary edematous inflammation, which is accompanied by massive recruitment of neutrophils to the lungs and onset of a turbulent proinflammatory cytokine storm within this organ [22,37]. Despite the similarity in morbidity and mortality, following pulmonary abrin and ricin intoxications in mice, the ability to protect mice against ricin and abrin intoxications by postexposure antibody-mediated treatment differs radically. In the case of lethal ricin intoxication, rabbitderived polyclonal anti-ricin antibody-based treatment of the mice at 24 h postexposure resulted in 34% survival rates [22]. When antibody treatment was administered at 48 h postexposure to ricin, protection was no more than marginal [38]. In sharp contrast, the administration of polyclonal anti-abrin antibodies to mice intranasally exposed to a lethal dose of abrin led to very high survival rates (~70–80%), even when the antibodies were applied as late as 72 h after intoxication [22]. This efficient protection by polyclonal antiabrin antibodies could not be attributed to the neutralization of a single subunit because specific antibodies against the A or B subunits of abrin were equally effective in protecting mice against pulmonary intoxication with chimeric reciprocal toxins harboring one of the subunits of abrin and the other of ricin [24]. These observations indicated that the difference in the protection conferred by anti-abrin and anti-ricin antibodies against abrin and ricin intoxications, respectively, is not related to the difference in the quality of the two antibody preparations. Therefore, in this study, we dissected the differences in abrin and ricin lung pathology following pulmonary exposure of mice to either of the toxins. We found that both toxins bound similarly to hematopoietic cells, especially to AMs and DCs, and triggered their early and persistent elimination from the lungs. These results are in agreement with an earlier study which showed that macrophages are the most sensitive cells to RIPs [39]. In opposition to cells of hematopoietic origin, the toxins differed in their efficiency of binding to parenchymal cells of the lungs. The binding of abrin to CD45- cells was considerably

less effective at all tested time points following intoxication. Consequently, following abrin intoxication, there was no direct damage found in the endothelial cell compartment, and the epithelial cell damage that, as in the case of ricin, was mostly confined to the alveolar epithelial type II cells was significantly lower than that observed following ricin intoxication. Supporting these results, we have previously shown that following intranasal intoxication, the in vivo catalytic performance of abrin, i.e., ribosomal depurination of pulmonary tissue, is significantly lower than that observed following ricin intoxication. In particular, the depurination levels of endothelial cells and pulmonary epithelial cells were markedly lower following abrin intoxication in comparison to ricin intoxication [40]. Furthermore, the lesser damage to lung parenchymal cells is in line with an older study that described the histopathology of the lungs in rats and showed that the appearance of apoptosis in the alveolar epithelium was far more marked following inhalation of ricin than abrin [19]. Interestingly, examination of the effect of antibody treatment against the two toxins on lung cell composition following exposure of mice to abrin and ricin showed significant reversion in the cells of hematopoietic origin, neutrophils and macrophages, after both intoxications [23]. However, neither the antibody-based treatment against abrin nor that against ricin conferred any beneficial influence on epithelial cells (data not shown). Since we did not find any repair of the epithelial population in the near term after antibody treatment, we estimate that the intensity of the epithelial damage has a direct effect on the survival of intoxicated and treated mice. As the epithelial damage in the lung is much more extensive after exposure to ricin compared to abrin, the protection ability of anti-ricin antibody treatment is limited.

One prominent hallmark of ricin-mediated pulmonary intoxication is the rapid and massive influx of neutrophils to the lungs [26–28]. This uncontrolled recruitment of neutrophils and the overwhelming activation in sterile inflammation, such as in the case of abrin or ricin intoxication, contributes to tissue damage by the release of proteinases, cationic polypeptides, cytokines and reactive oxygen species [41,42]. It has been previously shown that ricin does not bind neutrophils [25,40]. Similarly, we show in this study that infiltrating neutrophils, unlike other cells of hematopoietic origin, did not bind abrin following intoxication.

The dramatic influx over time of toxin-nonbinding neutrophils occurs mainly in the small capillaries spanning the alveolar network [43] and induces indirect lung damage by compromising the permeability of the alveolar–capillary barrier [44]. The kinetics and the extent of neutrophil influx to the lungs were found to be similar following abrin and ricin pulmonary intoxications. However, while the alveolar–capillary barrier integrity was compromised at early stages following ricin intoxication [27], lung permeability following abrin exposure was significantly increased only at later time points (48–72 h). This finding may stem from the relatively reduced levels of irreversible cellular damage inflicted by abrin. Alveolar–capillary barrier permeability is tightly regulated by the molecular interplay of intercellular junction molecules that span the gap between neighboring epithelial and endothelial cells. These junction molecules cooperate in maintaining tissue integrity to limit epithelial and endothelial permeability and to allow for just minimal leakage of fluids into the interstitial compartment [30,45]. We show in this study that in addition to reduced cellular damage, the insult to junction proteins, such as VE-cadherin and claudin 18, following abrin intoxication was less prominent, both at early and late stages, than the damage observed following ricin exposure. Moreover, the tight junction protein occludin, which is reduced at a later stage (48 h) following ricin intoxication [27], was not impaired at all post-abrin-exposure time points. These data may also account for the delayed hyperpermeability of the lungs post abrin exposure in comparison to ricin. However, the pronounced lung hyperpermeability at later time points after exposure to abrin may stem from robust recruitment of neutrophils at these time points and marked impairment in connexin 43 and claudin 5. In addition, although in contrast to ricin, no direct damage to endothelial cells was observed following abrin intoxication, collateral damage to these cells was discerned. In fact, we show in this study that the intensity of the damage to the

pulmonary vascular endothelial glycocalyx was comparable following intoxication with the two toxins.

In summary, we propose that the relatively superior performance of the anti-abrin antibody-based treatment and the ability to protect against abrin lethality, but not against ricin, when administered lately following intranasal intoxication in mice is due to the differences in the timing and intensity of the damage to the lung stroma inflicted by abrin and ricin.

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

#### *4.1. Animals*

Experiments were performed in accordance with Israeli law and approved by the Institutional Animal Care and Use Committee (IACUC) of the Israel Institute for Biological Research (Ness-Ziona, Iarael). Treatment of animals was in accordance with regulations outlined in the USDA Animal Welfare Act and the conditions specified in the National Institute of Health Guide for Care and Use of Laboratory Animals. Female CD-1 mice (27–32 g) were purchased from Charles River Laboratories Ltd., Margate, UK. Mice were housed in filter-top cages in an environmentally controlled room and maintained at 21 ± 2 ◦C and 55 ± 10% humidity. Lighting was set to mimic a 12/12 h dawn-to-dusk cycle. Mice were housed in a purpose-built animal holding facility for 4–8 days prior to the beginning of the experiment. Animals were allowed access to water ad libitum and 4% body weight food per day.

#### *4.2. Fluorescent Toxin Labeling and Intoxication*

Abrin and ricin were purified as previously described [22,23] and conjugated (1 mg) with the Alexa FluorTM 488 protein labeling kit (Molecular probes, Thermo Fisher Scientific, Waltham, MA, USA) according to the manufacturer's instructions. The cytotoxicity of labeled toxin (~5 dye/protein (mol/mol)) was determined in a cell-based assay developed in the past [46]. Briefly, labeled abrin or ricin was added to HEK-293 cell cultures, which secrete the enzyme acetylcholinesterase (AChE) in a constitutive manner. Secreted AChE was measured in the cell growth medium at 18 h postexposure, and activity levels were compared to those measured for nonlabeled toxin. We determined a ~2-fold reduction in the toxicity of the labeled toxins. Intranasal intoxication with toxin at a 2LD50 dose (unlabeled and labeled ricin or abrin, 7 and 14 μg/kg or 10 and 20 μg/kg, respectively) was applied (2 × 25 μL).

#### *4.3. Flow Cytometry*

Lungs were minced into small pieces and subjected to enzymatic digestion with 4 mg/mL collagenase D (Roche, Mannheim, Germany) for 2 h at 37 ◦C. The tissues were then meshed through a 40 μm cell strainer, and red blood cells were lysed with red blood cell lysis buffer (Sigma–Aldrich, Rehovot, Israel). For staining, cell suspensions were stained with CD45 (clone 30-F11), CD11b (M1/70), Ly6G (1A8), CD11c (N418), MHC class II (M5/114), Siglec F (S17007L), CD31 (390), CD326 (G8.8), T1α (8.1.1) and proSPC (Millipore, Temecula, CA, USA) followed by allophycocyanin (APC) donkey anti-rabbit IgG (Jackson ImmunoResearch, West Grove, PA, USA). All antibodies were purchased from Biolegend (San Diego, CA, USA) unless otherwise indicated. Neutrophils were identified as CD45high, Ly6Ghigh and CD11bhigh; AMs as CD45high, autofluorescent, Siglec Fhigh and CD11chigh; DCs as CD45high, Siglec Fneg and CD11chigh and MHC class IIhigh; endothelial cells as CD45neg and CD31high; epithelial cells as CD45neg and CD326high; alveolar epithelial cells type I as CD45neg and CD31neg, CD326high and T1αhigh; epithelial cells as CD45neg and CD326high; and alveolar epithelial cells type II as CD45neg and CD31neg, CD326high and proSPChigh cells. Flow cytometry was performed on a FACSCalibur (BD Biosciences, San Jose, CA, USA) and analyzed using FlowJo software v.10.8.0 (Tree Star, Ashland, OR, USA).

#### *4.4. Immunohistochemistry*

Lungs were collected and fixed in 4% buffered formaldehyde in PBS pH 7.2–7.4 (Bio Lab, Jerusalem, Israel) for 2 weeks. Sections of 5 μm were prepared after paraffin embedding using an RM 2255 microtome (Leica, Nussloch, Germany). Antigen retrieval was performed by incubation in Target Retrieval Solution (S1700, DAKO, Carpinteria, CA, USA, 30 min, 95 ◦C). After blocking in 5% BSA in PBS, slides were incubated (overnight, 4 ◦C) with purified anti-CD31 (390, Biolegend, San Diego, CA, USA), proSPC (Millipore, Temecula, CA, USA), podoplanin (T1α, 8.1.1, Biolegend), VE-cadherin (ab33168), claudin 5 (ab15106), connexin 43 (ab117843), occludin (ab31721) or claudin 18 (ab203563) (Abcam, Cambridge, MA, USA). Alexa Fluor 594- or 488-coupled donkey anti-rabbit or Alexa Fluor 594-coupled goat anti-Armenian hamster antibodies were used for detection (Molecular probes®, Thermo Fisher Scientific, Carlsbad, CA, USA). For nuclear staining, slides were mounted with Prolong® Gold antifade reagent containing DAPI (Molecular probes®, Thermo Fisher Scientific, Carlsbad, CA, USA). Analysis was performed using an LSM 710 confocal scanning microscope (Zeiss, Jena, Germany) equipped with the following lasers: argon multiline (458/488/514 nm), diode 405 nm, DPSS 561 nm and helium-neon 633 nm.

#### *4.5. Permeability Analysis*

Lung permeability was determined by the Evans blue dye (EBD) extravasation method as follows: EBD (7.5 mg/mL, Sigma–Aldrich, Rehovot, Israel) was injected intravenously into mice at a dose of 50 mg/kg and allowed to circulate for 1 h. Mice were then anesthetized, and the lungs were perfused by cutting the left atrium and flushing with 5 mL PBS through the right ventricle. The lungs were removed, and EBD was extracted by incubation of the tissues in 0.5 mL of formamide (Sigma–Aldrich, Rehovot, Israel) at 60 ◦C for 24 h. The EBD optical density in the supernatant was measured at 620 nm in a spectrophotometer (Molecular Devices, Sunnyvale, CA, USA), and the total amount of dye was calculated by means of a standard calibration curve.

#### *4.6. Analysis of Glycocalyx Shedding*

BALF was performed by flushing the lungs with 1 mL of PBS using a tracheal cannula. The BALF was centrifuged at 950× *g* at 4 ◦C for 10 min, and the supernatants were collected and tested for soluble heparan sulfate, hyaluronic acid and syndecan-1 levels using an LSBio (Seattle, WA, USA) Mouse Heparan Sulfate ELISA kit (LS-F39210), R&D Systems (Abingdon, UK) Porcine/Mouse Quantikine Hyaluronan Immunoassay kit (DHYAL0) and Diaclone (Besancon Cedex, France) Murine sCD138 (Syndecan-1) ELISA kit (860.090.096) according to the manufacturer's instructions.

#### *4.7. Statistical Analysis*

All statistical analyses were conducted with GraphPad Prism software (version 5.01, GraphPad Software Inc., La Jolla, CA, USA, 2007). Data are presented as the means ± SEMs. Significance was assessed by Student's t test, and for multiple comparisons, one-way analysis of variance (ANOVA) followed by Tukey's multiple comparisons test or two-way ANOVA with Bonferroni correction was used for planned comparisons. Differences were considered significant at *p* < 0.05. Points in graphs indicate individual mice.

**Author Contributions:** Conceptualization, A.S., T.S., C.K., Y.G. and R.F.; Formal analysis, A.S., C.K., Y.G. and R.F.; Investigation, A.S., Y.G., R.A., T.S., C.K. and R.F.; Data curation, A.S., Y.G., R.A., Y.E., T.S., C.K. and R.F.; Methodology, A.S., Y.G., R.A., Y.E., T.S., C.K. and R.F.; Visualization, A.S. and Y.E.; Validation, A.S., Y.G., T.S., C.K. and R.F.; Writing—original draft preparation, A.S. and R.F.; Writing—review and editing, A.S., Y.G., R.A., Y.E., T.S., C.K. and R.F.; Supervision, T.S. and C.K. All authors have read and agreed to the published version of the manuscript.

**Funding:** This study was supported by the Israel Institute for Biological Research.

**Institutional Review Board Statement:** The study was conducted in accordance with Israeli law and was approved by the Ethics Committee for Animal Experiments of the Israel Institute for Biological Research (project identification codes M-42-13, M-63-13, M-09-15 and M-44-15, approval dates 2 July 2013, 22 December 2013, 22 January 2015 and 23 June 2015, respectively).

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Acknowledgments:** We are grateful to the staff of the animal facilities of the Israel Institute of Biological Research.

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

#### **References**

