Phytotoxic Activity and Structure–Activity Relationships of Radicinin Derivatives against the Invasive Weed Buffelgrass (Cenchrus ciliaris)

Radicinin (1), is a fungal dihydropyranopyran-4,5-dione isolated together with some analogues, namely 3-epi-radicinin, radicinol, 3-epi-radicinol, and cochliotoxin (2–5), from the culture filtrates of the fungus Cochliobolus australiensis, a foliar pathogen of buffelgrass (Cenchrus ciliaris), an invasive weed in North America. Among the different metabolites 1 showed target-specific activity against the host plant and no toxicity on zebrafish embryos, promoting its potential use to develop a natural bioherbicide formulation to manage buffelgrass. These data and the peculiar structural feature of 1 suggested to carry out a structure-activity relationship study, preparing some key hemisynthetic derivatives and to test their phytotoxicity. In particular, p-bromobenzoyl, 5-azidopentanoyl, stearoyl, mesyl and acetyl esters of radicinin were semisynthesized as well as the monoacetyl ester of 3-epi-radicinin, the diacetyl esters of radicinol and its 3 epimer, and two hexa-hydro derivatives of radicinin. The spectroscopic characterization and the activity by leaf puncture bioassay against buffelgrass of all the derivatives is reported. Most of the compounds showed phytotoxicity but none of them had comparable or higher activity than radicinin. Thus, the presence of an α,β unsaturated carbonyl group at C-4, as well as, the presence of a free secondary hydroxyl group at C-3 and the stereochemistry of the same carbon proved to be the essential feature for activity.


Introduction
Buffelgrass (Cenchrus ciliaris or Pennisetum ciliare) is an important pasture grass in many semi-arid regions of the world but it is also an invasive weed in some areas of North America [1]. In the Sonoran Desert of southern Arizona it has infested thousands of acres of public and private lands, including Saguaro National Park and the Coronado and Tonto National Forests [2][3][4]. The increased fire frequency and intensity in the infested areas is negatively affecting the native species including the iconic saguaro cactus [5]. The main products currently used to manage buffelgrass are glyphosate and imazapyr. However, these broad-spectrum herbicides cause heavy damage to the non-target plants and have a negative environmental and ecological impact [6]. In the past decades the biological control has become an effective alternative to combat many weeds that invade natural systems [7,8]. In particular, the phytotoxins produced by weed pathogenic fungi are an efficient tool to design natural and safe bioherbicides [9]. Thus, Cochliobolus australiensis (recently classified as Curvularia tsudae) and Pyricularia grisea, two foliar pathogens that commonly occur on buffelgrass in the invaded North American range, were studied to evaluate their ability to produce phytotoxic metabolites that can potentially be used as natural herbicides against this weed. A total of 14 secondary metabolites belonging to different classes of natural compounds were purified from the in vitro cultures of these two pathogens [10][11][12]. When tested by leaf puncture assay on host plant at different concentrations, radicinin (1, Figure 1) and (10S,11S)-epi-pyriculol resulted to be the most promising compounds. Thus, their phytotoxic activity was also evaluated on non-host indigenous plants. Radicinin demonstrated high target-specific toxicity on buffelgrass, low toxicity to native plants and no teratogenic, sublethal, or lethal effects on zebrafish (Brachydanio rerio) embryos [13]. 1 is now under consideration for the development of target-specific bioherbicide to be used against buffelgrass, and a rapid and sensitive HPLC method for its quantification in complex mixtures was recently optimized in order to evaluate its production by different fungal strains and in different cultural conditions [14]. This manuscript reports the semisynthesis of some radicinin derivatives and the evaluation of their phytotoxic activity against buffelgrass. Furthermore, their phytotoxicity was also compared with that of the natural analogues 3-epi-radicinin, radicinol, 3-epi-radicinol, and cochliotoxin (2-5, Figure 1) and some of their derivatives (6-15, Figure 1) in order to obtain clues about the structure-activity relationship (SAR) of these compounds. non-target plants and have a negative environmental and ecological impact [6]. In the past decades the biological control has become an effective alternative to combat many weeds that invade natural systems [7,8]. In particular, the phytotoxins produced by weed pathogenic fungi are an efficient tool to design natural and safe bioherbicides [9]. Thus, Cochliobolus australiensis (recently classified as Curvularia tsudae) and Pyricularia grisea, two foliar pathogens that commonly occur on buffelgrass in the invaded North American range, were studied to evaluate their ability to produce phytotoxic metabolites that can potentially be used as natural herbicides against this weed. A total of 14 secondary metabolites belonging to different classes of natural compounds were purified from the in vitro cultures of these two pathogens [10][11][12]. When tested by leaf puncture assay on host plant at different concentrations, radicinin (1, Figure 1) and (10S,11S)-epi-pyriculol resulted to be the most promising compounds. Thus, their phytotoxic activity was also evaluated on non-host indigenous plants. Radicinin demonstrated high target-specific toxicity on buffelgrass, low toxicity to native plants and no teratogenic, sublethal, or lethal effects on zebrafish (Brachydanio rerio) embryos [13]. 1 is now under consideration for the development of target-specific bioherbicide to be used against buffelgrass, and a rapid and sensitive HPLC method for its quantification in complex mixtures was recently optimized in order to evaluate its production by different fungal strains and in different cultural conditions [14]. This manuscript reports the semisynthesis of some radicinin derivatives and the evaluation of their phytotoxic activity against buffelgrass. Furthermore, their phytotoxicity was also compared with that of the natural analogues 3-epi-radicinin, radicinol, 3-epi-radicinol, and cochliotoxin (2-5, Figure 1) and some of their derivatives (6-15, Figure 1) in order to obtain clues about the structure-activity relationship (SAR) of these compounds.  Figure 1. The structures of radicinin (1), 3-epi-radicinin (2), radicinol (3), 3-epi-radicinol (4), cochliotoxin (5), radicinin derivatives (6-10, 14, and 15), and the acetyl derivatives of 3-epi-radicinin, radicinol, and 3-epi-radicinol (11-13, respectively).

Results and Discussion
The buffelgrass pathogenic fungus C. australiensis was grown by fermentation and the natural compounds 1-5 were isolated from potato dextrose broth (PDB) cultures according to the procedures previously published [10,11]. The purity of 1-5 was >98%, as checked by 1 H-NMR and LC-MS.

Results and Discussion
The buffelgrass pathogenic fungus C. australiensis was grown by fermentation and the natural compounds 1-5 were isolated from potato dextrose broth (PDB) cultures according to the procedures previously published [10,11]. The purity of 1-5 was >98%, as checked by 1 H-NMR and LC-MS.
To investigate the structure activity relationship for radicinin as a target-specific phytotoxin against buffelgrass, seven derivatives (6-10 and 14-15, Figure 1) were prepared as reported in detail in the experimental part. The acetyl derivatives of 3-epi-radicinin, radicinol, and 3-epi-radicinol 11-13 were also prepared. All the derivatives were characterized as described in Materials and Methods in detail, and their 1 H-NMR data are reported in Tables 1 and 2.  Radicinin (1) by reaction with 4-bromobenzoyl chloride yielded its corresponding p-bromobenzoyl ester (6). Its 1 H-NMR spectrum differed from that of 1 for the presence of the typical signals pattern of the aromatic para-disubstituted residue, appearing as two doublets at δ 7.94 and 7.65 (J = 8.6 Hz), and for the downfield shift (∆δ 1.52) at δ 5.52 of the signal of H-3. These data were very similar to those already reported by Robeson et al. [15]. As a stronger evidence of the derivatization, the ESI-MS spectrum showed the typical signals because of the presence of 79 Br and 81 Br isotopic peaks, at m/z 419 [M + H] + and 421 [M + 2 + H] + , respectively.
Radicinin by esterification with 5-azidopentanoic acid was converted into the corresponding 5-azidopentanoyl derivative (7). Its 1 H-NMR spectrum differed from that of radicinin for the downfield shift (∆δ = 1.25) of H-3 at δ 5.25 and for the presence of the signals pattern typical of 5-azidopentanoyl residue resonating as two triplets at δ 3.35 (J = 6.5 Hz) and 2.53 (J = 7.2 Hz) due to CH 2 -5 and CH 2 -2 , and two multiplets at δ 1.80−1.72 and 1.68−1.71 due to CH 2 -3 and CH 2 -4 . Further confirmation was obtained by the stretching of the -N 3 bond in the IR spectrum, with a signal at 2097 cm −1 and by the ESI-MS spectrum, which showed both the dimeric sodiated [2M + Na] + and protonated [M + H] + forms at m/z 745 and 362, respectively.
1 by reaction with stearoyl chloride afforded the corresponding acyl derivative (8). Its 1 H-NMR spectrum, compared with that of radicinin, differed for the downfield shift (∆δ = 1.27) of H-3 at δ 5.27 and showed typical signals of the stearoil residue appearing as two triplets at δ 2.44 (J = 7.5 Hz) and 0.90 (J = 6.7 Hz) due to CH 2 -2 and Me-18 and the presence at δ 2.0-1.0 of the complex multiplet due to the protons of the residual 15 CH 2 groups. The ESI-MS spectrum showed the protonated [M + H] + form at m/z 517.
Radicinin by reaction with mesyl chloride in pyridine, afforded the corresponding mesyl ester (9). Its 1 H-NMR spectrum, compared to that of 1 showed the downfield shift of H-3 (∆δ = 1.01) appearing as a doublet (J = 12.4 Hz) at δ 5.01 and the singlet of the mesyl group at δ 3.40. A further confirmation was obtained by the ESI-MS spectrum, which showed the dimeric sodiated [2M + Na] + and the protonated [M + H] + forms at m/z 651 and 315, respectively.
Radicinin (1) was acetylated by usual reaction with Ac 2 O and pyridine to yield the corresponding 3-O-acetylderivative (10). Its 1 H-NMR differed from that of 1 essentially for the downfield shift of H-3 (∆δ = 1.28) appearing as singlet at δ 5.28 and the singlet of the acetyl group at δ 2.23. These data were very similar to those previously reported [16]. Furthermore, its ESI-MS spectrum showed the dimeric sodiated [2M + Na] + and protonated [M + H] + forms at m/z 579 and 279, respectively.
The ten hemisynthetic derivatives (6)(7)(8)(9)(10)(11)(12)(13)(14)(15) were tested by the buffelgrass leaf puncture assay at 2.5 × 10 −3 M as reported in the Materials and Methods section in detail. Their activity was evaluated in comparison with that showed by the natural metabolites (1-5) previously reported by Masi et al. [13]. Overall, five derivatives (6, 8, 12, 13, and 15) produced no necrosis and were thus completely nontoxic to buffelgrass, while the other five compounds (7, 9, 10, 11, and 14) were moderately toxic ( Figure 2). A very important factor for the activity appear to be the carbonyl group of the dihydro γ-pyrone as the phytotoxicity was lost in 3. This was also confirmed by the expected lacking of activity of radicinol diacetyl derivative (12). The stereochemistry at C-3 also plays a significant role to impart activity as was with the strong reduction observed by testing 3-epi-radicinin (2). This was confirmed by the very low activity of its acetyl derivative 11 and the total loss of activity observed by testing the diacetyl derivatives (12 and 13) of radicinol and 3-epi-radicinol. The hydrogenation of 1 generated 14 which showed the absence of α,β unsaturated carbonyl group explaining the strong loss of activity. The double bond of the side chain at C-7 also plays a role to impart activity as demonstrated by the reduction of phytotoxicity observed by testing 5 and confirmed by the decrease of the activity of 14. The acyl derivatives of 1 showed a strong or total loss of activity, which was only in part retained by the acetyl and the mesyl derivatives (9 and 10). These compounds are probably more easily hydrolysable in comparison to the other compounds (6, 7, and 8) as their conjugated bases are more stabilized for resonance. A very important factor for the activity appear to be the carbonyl group of the dihydro γ-pyrone as the phytotoxicity was lost in 3. This was also confirmed by the expected lacking of activity of radicinol diacetyl derivative (12). The stereochemistry at C-3 also plays a significant role to impart activity as was with the strong reduction observed by testing 3-epi-radicinin (2). This was confirmed by the very low activity of its acetyl derivative 11 and the total loss of activity observed by testing the diacetyl derivatives (12 and 13) of radicinol and 3-epi-radicinol. The hydrogenation of 1 generated 14 which showed the absence of α,β unsaturated carbonyl group explaining the strong loss of activity. The double bond of the side chain at C-7 also plays a role to impart activity as demonstrated by the reduction of phytotoxicity observed by testing 5 and confirmed by the decrease of the activity of 14.
The acyl derivatives of 1 showed a strong or total loss of activity, which was only in part retained by the acetyl and the mesyl derivatives (9 and 10). These compounds are probably more easily hydrolysable in comparison to the other compounds (6, 7, and 8) as their conjugated bases are more stabilized for resonance.
The five hemisynthetic derivatives (7, 9, 10, 11, and 14) that showed toxic activity on buffelgrass leaves at 2.5 × 10 −3 M were also tested at a lower concentration of 10 −3 M (Figure 3), confirming the results obtained at higher concentration.
The five hemisynthetic derivatives (7, 9, 10, 11, and 14) that showed toxic activity on buffelgrass leaves at 2.5 × 10 −3 M were also tested at a lower concentration of 10 −3 M (Figure 3), confirming the results obtained at higher concentration.

Fungal Strains
Cochliobolus australiensis (LJ-4B) strains used in this study were isolated from diseased buffelgrass tissue collected in Saguaro National Monument, Arizona, AZ, USA, in autumn 2014 and near La Joya, Hidalgo County in south Texas, USA, in September 2014, respectively.

Isolation of Fungal Metabolites
The fungal metabolites 1-5 were isolated from in vitro PDB (potato dextrose broth) cultures of C. australiensis according to procedures previously reported [10,11].

Fungal Strains
Cochliobolus australiensis (LJ-4B) strains used in this study were isolated from diseased buffelgrass tissue collected in Saguaro National Monument, Arizona, AZ, USA, in autumn 2014 and near La Joya, Hidalgo County in south Texas, USA, in September 2014, respectively.

Isolation of Fungal Metabolites
The fungal metabolites 1-5 were isolated from in vitro PDB (potato dextrose broth) cultures of C. australiensis according to procedures previously reported [10,11].

Leaf Puncture Bioassays
The ten hemisynthetic derivatives (6)(7)(8)(9)(10)(11)(12)(13)(14)(15) were first assayed at 2.5 × 10 −3 M for phytotoxicity on the leaves of buffelgrass (Cenchrus ciliaris) and those active at this concentration were then bioassayed at 10 −3 M. Compounds were first dissolved in MeOH (final concentration 4%) and stock solutions at the two concentrations using sterile distilled water were then prepared. An incision of ca. 3 mm was made on the adaxial surface of each leaf section of 3 cm with an insulin needle. The leaf sections were placed in groups of six on the surface of a water-saturated filter paper in each of the four petri dishes. A total of five leaf sections in each petri dish were tested with the solution containing the compound, while one leaf section was used as a negative control (4% MeOH only). A droplet (10 µL) of the appropriate solution was applied over each needle incision using a micropipette. The dishes were sealed with parafilm and incubated at 24 • C for 3 days in a temperature-regulated chamber under a photoperiod of 14-10 h (light/dark). After 3 days of treatment, necrotic lesion development was evaluated by removing the petri dish cover, placing a glass disc on the leaf sections to flatten them into a single plane, and photographing each dish with its leaf sections. Each acquired image was then analyzed with the software ImageJ to measure the necrotic area caused by the solution.

Statistical Analyses
Statistical analyses were carried out using the GraphPad Prism 8 software (GraphPadSoftware, San Diego, CA, USA). Data were represented as the mean ± standard deviation and analyzed for statistical significance using ordinary one-way or two-ways analysis of variance and multiple comparisons. For all test, p < 0.5 was considered to indicate a statistically significant difference.

Conclusions
These results obtained in this study demonstrated that the α,β unsaturated carbonyl group at C-4 and the stereochemistry at C-3 are important structural features to impart phytotoxicity. Furthermore, the unsaturation of propenyl side chain also play a role to impart activity.
Thus, radicinin appears to be the most active compound suitable to develop a target-specific bioherbicide for buffelgrass control. Considering the very low production of this compound by different strains of the fungus C. australiensis [14] and the difficulties to scale-up the cultures via a fermenter, a suitable alternative appears to be its total enantioselective synthesis.
Author Contributions: All the authors contributed in the designing of the study, the analysis of the data, and writing of the manuscript. M.M. performed the conceptualization, acquired the funding, and wrote a draft of the manuscript; F.F. performed the bioassays and M.M and AE analyzed the data; A.C., M.C., S.M., and A.E. finalized the draft and revised the manuscript.
Funding: This research was funded by Programme STAR 2017, financially supported by UniNA and Compagnia di San Paolo grant number E62F16001250003 and by BBCA onlus Foundation, Rome, Italy.