The identification of small molecule inhibitors of the plant inositol phosphorylceramide synthase which demonstrate herbicidal activity

Resistance to 157 different herbicides and 88% of known sites of action has been observed, with many weeds resistant to two or more modes. Coupled with tighter environmental regulation, this demonstrates the need to identify new modes of action and novel herbicides. The plant sphingolipid biosynthetic enzyme, inositol phosphorylceramide synthase (IPCS), has been identified as a novel, putative herbicide target. The non-mammalian nature of this enzyme offers the potential of discovering plant specific inhibitory compounds with minimal impact on animals and humans, perhaps leading to the development of new non-toxic herbicides. The best characterised and most highly expressed isoform of the enzyme in the model-dicot Arabidopsis, AtIPCS2, was formatted into a yeast-based assay which was then utilized to screen a proprietary library of over 11,000 compounds provided by Bayer AG. Hits from this screen were validated in a secondary in vitro enzyme assay. These studies led to the identification of a potent inhibitor that showed selectivity for AtIPCS2 over the yeast orthologue, and activity against Arabidopsis seedlings. This work highlighted the use of a yeast-based screening assay to discover herbicidal compounds and the status of the plant IPCS as a novel herbicidal target.

This divergence in sphingolipid biosynthesis has been exploited to investigate the protozoal IPCS as a therapeutic target for the Neglected Tropical Diseases, Chagas disease [12][13][14] and leishmaniais [15][16][17][18] . In plants, the activity of IPCS was first characterized in Phaseolus vulgaris 19 and its role as a negative regulator of programmed cell death in plants was validated in Arabidopsis thaliana 20 and Eucalyptus grandis 21 . In Oryza sativa, IPCS has been shown to play a role in plant response to abiotic stress, particularly in response to drought, cold and salinity 22 .
Despite the fact that hundreds of herbicides are widely used, these only exhibit 25 modes of action. In fact, merely 6 modes of action, targeting 5-enolpyruvylshikimate-3-phosphate (EPSP) synthase, acetolactate synthase (ALS), photosystem (PS) II, synthetic auxins, acetyl CoA carboxylase (ACCase) and cell division, acount for 75% of the herbicide market 23 . It has been over 30 years since a herbicide with a new mode of action was introduced onto the market and, with the growing problem of herbicide resistance 24 and the destabilizing effect of climate change on crop yield 25 , it is now necessary to identify new herbicidal modalities to ameliorate the challenge of feeding a rapidly increasing global population set to reach 9-10 billion in 2050 26 .
As previously reported [20][21][22] , inhibition of the plant IPCS would lead to a buildup of the enzyme substrate, the Programmed Cell Death (PCD; apoptosis) mediator phytoceramide 27 . The functional divergence of IPCS from the equivalent mammalian enzyme, sphingomyelin synthase (SMS), could allow the identification of specific, non-toxic inhibitors. This possibility has, to date, lead to the identification of 5 potent inhibitors of the fungal IPCS (aureobasidin A 28 (AbA), khafrefungin 29 , rustimicin 30 , pleofungin 31 and haplofungin 32 ) with low nano-molar IC 50 values against Saccharomyces cerevisiae. However, currently, no inhibitor of the plant orthologue has been identified.
In this study, the well characterized Arabidopsis thaliana enzyme AtIPCS2 20 , the most highly expressed of the 3 IPCS isoforms 33 , was used to complement S. cerevisiae lacking AUR1. The yeast utilized was engineered to enhance compound sensitivity through reduced expression of several efflux pumps, and thereby allow efficient hit identification in a cell-based high throughput screening (HTS) assay used for 11,440 bioactive compounds. A secondary enzyme-based assay facilitated the validation of hits as inhibitors of the enzyme, and enabled comparison of their activity against the yeast orthologue, Aur1p. This allowed the identification of hits that exhibited selectivity for IPCS2 from A. thaliana. The most potent selective compound was tested in vivo against seedlings and demonstrated herbicidal activity.

Results
Primary high throughput screening using a yeast-based assay. Fungi such as Saccharomyces cerevisiae possess multiple genes linked to pleiotropic drug resistance, including those encoding a range of ATPbinding cassette (ABC) transporters and the transcription factors required for their expression 34 . These extrusion pumps can be over-expressed in response to drug treatment, leading to decreased intracellular drug concentrations and subsequent drug resistance 35 , while multiple deletions of these functions render yeast cells significantly more sensitive to a range of toxic compounds including antifungal agents used in agriculture and medicine 8 . To increase the sensitivity of the yeast-based assay platform, an S. cerevisiae strain was utilised that lacked PDR1, PDR3, PDR16 and PDR17. This combination of pdr deletions was shown to confer significant hypersensitivity to a range of compounds ( Supplementary Information 1). PDR1 36 and PDR3 37 encode paralogous Zn(II) 2 Cys 6 zinc finger regulators, which control the transcription of ABC drug efflux pump-encoding genes including PDR5 38,39 , SNQ2 40 , PDR10 41 , PDR15 41 and YOR1 42 through binding to cis-acting PDREs (pleiotropic drug resistance elements) 40,41,43,44 . PDR16 and PDR17 encode a pair of paralogous phosphatidylinositol transport proteins that also confer drug hypersensitivity when deleted 45 .
In the quadruple pdr1∆ pdr3∆ pdr16∆ pdr17∆ strain, AUR1 was deleted and replaced by a HIS3 selectable marker, with growth supported by expression of the essential AUR1 gene from the plasmid pRS316-AUR1 under uracil selection. In this background, galactose-inducible expression of AtIPCS2 from plasmid pESC-LEU was found to complement loss of pRS316-AUR1 when the yeast were cultured in the presence of 5-fluoroorotic acid. This made the yeast dependent upon the presence of galactose for growth, thus demonstrating dependence on the expression of the plant enzyme ( Fig. 1, Supplementary Information 2) [20][21][22] . Assay of microsomal extracts from the complemented yeast demonstrated IPCS activity in vitro and confirmed that the plant activity is insensitive to the fungal Aur1p inhibitor aureobasidin A (AbA) 28,33 (Fig. 2).
Yeast complemented with the well characterised AtIPCS2, and an AUR1 control, were subsequently formatted into a 96-well plate. Following statistical validation by calculation of Z factor 46 in the presence of positive (cycloheximide) and negative (DMSO) controls, the assay was used in HTS of a focused library of 11,440 bioactive compounds. All assay plates were required to have a calculated Z factor ≥0.5 for the data to be progressed. Following in duplicate screening at 10 µM against AtIPCS2 complemented yeast and the AUR1 control, www.nature.com/scientificreports www.nature.com/scientificreports/ compounds exhibiting ≥80% inhibition and ≥50% selectivity for AtIPCS2 were taken forward. After eliminating false positives (non-reproducible hits; 2.6%), 106 target directed hits were identified, a hit rate of 0.9% (Fig. 3, Supplementary Information 3). It is notable that a significant minority of compounds increased yeast proliferation and that this phenotype was more profound in the AtIPCS2 complemented yeast (negative inhibition; Fig. 3), whilst this is an interesting observation these were not analysed further. Dose response analyses (50 µM to 68 nM), using the same assay platform, demonstrated that the majority of the inhibitory compound hits (89 of 106) had an IC 50 of less than 10 µM (Supplementary Information 4).
Secondary screening using an in vitro biochemical assay. In the secondary screening stage the previously described microsomal-based in vitro IPCS assay was adapted and utilised 17 . Initially, all 106 selective hits from the primary screen were tested, in duplicate, at 10 µM. 16 compounds which, reproducibly, showed ≥30% inhibition were carried forward for in triplicate dose response (100 µM to 46 nM) analyses and IC 50 determination against AtIPCS2 and, as a control, AUR1. All were active to some degree against the Arabidopsis enzyme, whilst none showed inhibition of the fungal orthologue, demonstrating that selective AtIPCS2 inhibitors had been identified. 4 compounds demonstrated IC 50 values < 10 µM (Compound 1, 4.02 µM; 2, 4.75 µM; 3, 8.41 µM; and 4, 9.84 µM; Supplementary Information 5). The structures of compounds 2-4 are withheld due to intellectual property reasons, leaving the most active (Compound 1, a phenylamidine carrying an acetonitrile functional group) to be taken forward (Fig. 4). The structural integrity of Compound 1 was confirmed using mass spectrometry and 1 H and 13 C spectroscopy which showed that it was a 3:2 mixture of E:Z amidine isomers (see Supplementary Information 6).
In vivo screening. In vivo testing of the phenylamidine Compound 1 was undertaken against Arabidopsis seedlings grown on agar. Dose response analyses (Fig. 5) showed that treatment restricted growth and led to purple leaf patches at 11 µM and above. Examination of treated 7 day old seedlings grown on agar containing 10 and 40 µM of Compound 1 showed plants with clear purple patches associated with anthocyanin biosynthesis in response to stress 47 , and an absence of lateral root development compared to the DMSO control (Fig. 6).

Discussion
With herbicide resistance increasing 24 and climate change effecting on crop yield 25 , the need to identify new herbicide targets and lead molecules to address these challenges is pressing. One major hurdle to overcome in this search for a new herbicide is to ensure identified chemicals have acceptable toxicity profiles which are safe to the user and the environment 48 . The divergence in the sphingolipid biosynthetic pathway between mammals and plants, where the former produce SM and the latter IPC 20,33 , may present an opportunity to identify molecules with such a profile.
Following the recent publication of our successful HTS campaign against a protozoan IPCS 49 , this study is the first report of HTS for inhibitors of an enzyme in the plant sphingolipid synthetic pathway, the non-mammalian AtIPCS2 -the most highly expressed and best characterised isoform in the model dicot Arabidopsis, which catalyses the synthesis of IPC 20,33 . The role of this enzyme in phytoceramide homeostasis 20,33 and therefore PCD 50 , coupled with the product, IPC, functioning as the precursor for the synthesis of glycosylinositol phosphorylceramide (GIPC; 25% of plasma membrane lipid 51 ), makes IPCS an attractive target for the discovery of new, non-toxic, herbicidal agents. However, given the multi-transmembrane nature of the enzyme 20,33 assay development is www.nature.com/scientificreports www.nature.com/scientificreports/ challenging. Therefore, to facilitate HTS, we developed a novel cell-based assay utilising an AtIPCS2 complemented S. cerevisiae strain lacking 4 extrusion pumps linked to pleiotropic drug resistance (PDR1, PDR3, PDR16 and PDR17) to increase sensitivity, and utilised this system to screen a library of 11,440 bioactive compounds.   www.nature.com/scientificreports www.nature.com/scientificreports/ Counter screening against Aur1p (the yeast orthologue) formatted in the same assay yielded 106 selective hits, of these 4 demonstrated IC 50 values < 10 µM in a secondary in vitro enzyme assay and minimal activity against yeast Aur1p (>50 µM). The most active was a phenylamidine, Compound 1 (IC 50 < 5 µM), which has been patented by Bayer as a fungicide 52 . Previous phylogenic analyses 33 have shown that the three IPCS isoforms in Arabidopsis are closely related. Further, focused, sequence analyses demonstrated that whilst AtIPCS2 orthologues are highly conserved within the monocots and eudicots, there is distance between the two clades (see Supplementary  Information 7). This indicated that selective inhibition of the enzyme (for example in a weed species) maybe feasible. Future studies should examine the selectivity of Compound 1 for AtIPCS2 over the other two isoforms and other plant orthologues to establish selectively in Planta.
The phenylamidines were first identified in the 1960s as pesticides for the control of plant fungal pathogens 52 and specific variants have subsequently been patented for use as herbicides 53 . In vivo screening of the identified phenylamidine, Compound 1, against wild type Col-0 Arabidopsis seedlings demonstrated dose dependent effects with decreased lateral root development, and distinctive purple leaf patches associated with anthocyanin biosynthesis in response to stress 47 . The mode of action of herbicidal phenylamidines have not been published, but the phenotypic effects reported here for Compound 1 are consistent with those expected for an IPCS inhibitor.
In conclusion, using a novel HTS approach the first inhibitor (Compound 1 -a phenylamidine) of plant IPCS was identified and shown, in vivo, to induce the plant stress response. This low molecular weight compound is ideal for further development towards use in agriculture, and further studies are planned to investigate this possibility.
In vivo screening, Arabidopsis seedlings. A. thaliana (Col0) seedlings were grown for 10 days on 0.8% Murashige and Skoog (MS) agar and then transferred to 1.2% MS agar containing compounds at the desired concentrations or DMSO as a control. Plants were grown at 20 °C under 16 hour day/8 hour night photoperiod.