Common Molecular Targets of a Quinolone Based Bumped Kinase Inhibitor in Neospora caninum and Danio rerio

Neospora caninum is an apicomplexan parasite closely related to Toxoplasma gondii, and causes abortions, stillbirths and/or fetal malformations in livestock. Target-based drug development has led to the synthesis of calcium-dependent protein kinase 1 inhibitors, collectively named bumped kinase inhibitors (BKIs). Previous studies have shown that several BKIs have excellent efficacy against neosporosis in vitro and in vivo. However, several members of this class of compounds impair fertility in pregnant mouse models and cause embryonic malformation in a zebrafish (Danio rerio) model. Similar to the first-generation antiprotozoal drug quinine, some BKIs have a quinoline core structure. To identify common targets in both organisms, we performed differential affinity chromatography with cell-free extracts from N. caninum tachyzoites and D. rerio embryos using the 5-aminopyrazole-4-carboxamide (AC) compound BKI-1748 and quinine columns coupled to epoxy-activated sepharose followed by mass spectrometry. BKI-binding proteins of interest were identified in eluates from columns coupled to BKI-1748, or in eluates from BKI-1748 as well as quinine columns. In N. caninum, 12 proteins were bound specifically to BKI-1748 alone, and 105 proteins, including NcCDPK1, were bound to both BKI-1748 and quinine. For D. rerio, the corresponding numbers were 13 and 98 binding proteins, respectively. In both organisms, a majority of BKI-1748 binding proteins was involved in RNA binding and modification, in particular, splicing. Moreover, both datasets contained proteins involved in DNA binding or modification and key steps of intermediate metabolism. These results suggest that BKI-1748 interacts with not only specific targets in apicomplexans, such as CDPK1, but also with targets in other eukaryotes, which are involved in common, essential pathways.


Introduction
Apicomplexan parasites, in particular, the genera Plasmodium, Cryptosporidium, Eimeria, Babesia, Theileria, Sarcocystis, Toxoplasma, Besnoitia and Neospora, are pathogens causing serious disease in humans as well as in animals. In cattle, sheep or other ruminants, Neospora caninum, closely related to Toxoplasma gondii, causes abortion, stillbirths or birth of weak offspring [1,2]. Since neither vaccines nor therapeutic agents against neosporosis are currently available [3,4], there is an increasing interest in novel compounds potentially useful for the treatment of this disease.
Novel chemotherapies against neosporosis could be generated via the repurposing of known drugs, e.g., against malaria [5][6][7] or by de novo target-based drug development. An example of such targets are calcium-dependent protein kinases (CDPKs) with homologs in plants, apicomplexans and other phyla, but not in animals. Consequently, during the last decade, CDPKs, in particular, CDPK1, have been intensively investigated as targets for drug development against Toxoplasma gondii [8] and other related apicomplexans [9]. These studies have led to the development of bumped kinase inhibitors (BKIs) of CDPK1 [10,11]. CDPK1 is crucially involved in invasion of host cells and egress [12].
In studies using the mouse model, the pyrazolopyrimidine (PP) BKI-1294 has yielded promising results against vertical transmission of neosporosis [13,14] and toxoplasmosis [15] in mice, as well as against toxoplasmosis in sheep [16]. Another PP compound, BKI-1553, was also shown to exhibit protective effects in mice and inhibited vertical transmission and abortion in N. caninum infected pregnant sheep [17]. Other BKIs were, however, less promising. In particular, the 5-aminopyrazole-4-carboxamide (AC) BKI-1517 showed good therapeutic efficacy against neosporosis in adult animals and protection from vertical transmission but was shown to have detrimental effects on fertility in mice [18]. Another BKI with promising efficacy in vitro and in vivo is the AC compound BKI-1748 [19]. This BKI had a profound impact on vertical transmission of N. caninum in pregnant mice when applied at 20 mg/kg/day for five days, but fertility in mice was dramatically reduced when applied at 50 mg/kg/day for five days.
In vitro studies using N. caninum infected fibroblasts showed that several BKIs, including BKI-1294, BKI-1553 and BKI-1748, did not act parasiticidally. Rather, exposure of intracellular tachyzoites to these compounds induced the formation of multinucleated complexes, which were composed of newly formed zoites devoid of the outer plasma membrane, and were incapable of separating from each other and forming infective tachyzoites [13,14,19]. These findings suggest that BKIs could act on other targets besides CDPK1.
In addition, several BKIs, including BKI-1748, were shown to cause embryonic malformations in a zebrafish (Danio rerio) model when applied during the first 96 h postfertilization at concentrations of 20 µM and above [20]. BKI-1517 and BKI-1553 also exerted embryonic malformations in zebrafish, although already at 0.2 and 5 µM, respectively [20]. These BKIs share a quinoline core (see structures in [20]), which may at least be partially responsible for the side effects of these compounds. Quinine, one of the first antiprotozoal drugs, has long been known to affect pregnancy [21,22], as discussed elsewhere [23]. Consequently, we hypothesize that BKI-1748 could interact with common targets in apicomplexans and animals, and that this quinoline motif could play a role. To investigate our hypothesis, we performed differential affinity chromatography followed by mass spectrometry with BKI-1748 and quinine ( Figure 1) on cell-free extracts from N. caninum tachyzoites and D. rerio embryos.
Here, we present differential affinity chromatography (DAC) results showing that common BKI-1748 binding proteins were identified in both organisms.

Scanning Electron Microscopy (SEM) of Neospora caninum and Danio rerio Treated with BKI-1748
When human foreskin fibroblasts were infected with N. caninum tachyzoites in the absence of BKI-1748, numerous tachyzoites were formed and egressed from the host cells.
In the presence of BKI-1748, parasites formed larger complexes that had to be physically released from their host cells in order to visualize them for SEM ( Figure 2). After 96 h post-fertilization, untreated D. rerio embryos exhibited a normal phenotype, as expected for this stage of development, as depicted in Figure 3. When treated with BKI-1748, D. rerio embryos showed distinct malformations, such as an enlarged thorax (at 20 µM) or even a burst of the thorax (at 40 µM). Moreover, the eye lost its characteristic morphological appearance upon BKI-treatment (

DAC Proteomes of Neospora caninum and Danio rerio
Mass spectrometry analysis of the proteomes obtained after DAC of cell-free extracts of N. caninum tachyzoites resulted in the identification of 8089 unique peptides matching to 1162 proteins. The complete dataset is given in Supplementary Table S1, which is available online. The corresponding numbers for the proteomes obtained by DAC of D. rerio embryos were 9381 unique peptides matching to 1202 proteins. The complete dataset is given in Supplementary Table S2. In both datasets, the protein intensity distributions were almost equal for all column eluates ( Figure 5).  Table S2 (D. rerio) as calculated by the Top3, LFQ and iBAQ algorithms. Cell-free extracts were prepared and subjected to differential affinity chromatography on mock (green), quinine (orange), or BKI-1748 (blue) columns followed by mass spectrometry as described in the Materials and Methods Section.
Unbiased analysis of the dataset by principal component analysis (PCA) demonstrated that the differential affinoproteomes eluted from mock, quinine and BKI-1748 columns were located in non-overlapping clusters separated by both principal components. This was true for both organisms and with both Top3 and LFQ algorithms ( Figure 6).
After the subtraction of proteins binding to the mock columns, the affinoproteome of N. caninum comprised 424 proteins specifically binding to BKI-1748 or quinine columns or both, the affinoproteome of D. rerio comprised 357 proteins, as depicted in Figure 7.

Specific BKI-1748 Binding Proteins in N. caninum and D. rerio
As shown in Figure 7, only 12 N. caninum proteins could be identified to be binding specifically to BKI-1748 (Table 1). NcCDPK1 (NCLIV_011980), the target, for which BKI-1748 and related compounds had been designed, was not identified as a protein specifically binding to BKI-1748 but was also detected in eluates from mock and quinine columns. The respective LFQ values were 22 × 10 6 (BKI-1748), 4 × 10 6 (quinine) and 9 × 10 6 (mock; Supplementary Table S1). Table 1. List of the 12 proteins from Neospora caninum specifically binding to BKI-1748, as identified by differential affinity chromatography followed by mass spectrometry. See Supplementary Table S1 for the full dataset. The relative abundance (rAbu) is based on iBAQ, so that the sum or rAbu is 1,000,000 for every column. The ORF numbers from ToxoDB are provided.

ORF Number
Annotation rAbu Valyl-tRNA synthetase, mitochondrial, related 29 NCLIV_066970 Putative enoyl-acyl carrier reductase 17 NCLIV_051300 Nifu-pending protein, related 14 NCLIV_022400 Hypothetical protein 9 By far, the most abundant BKI-1748 binding protein is a 260 amino acid hypothetical protein encoded by ORF NcLIV_020480 ( Figure 8A). The closest homologs are U1 70 kDa small nuclear ribonucleoproteins conserved amongst various phyla. Since the protein encoded by ORF NcLIV_020480 is much shorter, the homologies only extend to the Nterminal sequences of these ribonucleoproteins ( Figure 8B).
The closest homolog for structural modeling was the template 6qx9.6.A with 51.30% sequence identity corresponding to the U1 small nuclear ribonucleoprotein 70 kDa from the structure of a human fully-assembled precatalytic spliceosome pre-B complex [24], presented in Figure 8C. Interestingly, the second most abundant protein NCLIV_069960 encoding an RNAbinding protein, and two other BKI-1748 binding proteins, an RNA polymerase, and a t-RNA-synthase, were related to transcriptional and translational processes.
Moreover, alveolin1, member of a family of proteins uniquely found in the phylum Alveolata [25], was amongst the proteins binding specifically to BKI-1748.
The D. rerio pull-down yielded 13 proteins specifically binding to BKI-1748. By far, the most abundant protein was A0A140LGT9_DANRE encoding a 155 amino acid protein with a predicted signal peptide at its N-terminus (amino acids . The processed peptide chain is homologous to leucocyte-derived chemotaxin 2 from other teleost species; the second most abundant binding protein was syntenin ( Table 2). Table 2. List of the 13 proteins from Danio rerio embryos specifically binding to the BKI-1748 column, as identified by differential affinity chromatography followed by mass spectrometry. See Supplementary Table S2 for the full dataset. The relative abundance (rAbu) based on iBAQ, so that the sum or rAbu is 1,000,000 for every column.

Uniprot Number
Annotation rAbu Moreover, proteins involved in RNA binding and modification, namely splicing factor 1, a microsomal ribsosomal protein, and a pre-mRNA processing factor were identified.

Proteins Binding to Both BKI-1748 and Quinine Columns
Besides proteins binding specifically to BKI-1748, proteins binding to both BKI-1748 and quinine columns are of interest. In eluates from N. caninum tachyzoites, 105 proteins binding to both columns were identified (Supplementary Table S3). Within this subset, the most abundant protein was the 180 amino acid polypeptide encoded by ORF NCLIV_068520, annotated as a MIC5 homolog by the ToxoDB. In T. gondii, the microneme protein MIC5 is secreted onto the parasite surface during invasion and acts as an inhibitor of the proteolytic processing of microneme and dense granule proteins secreted by tachyzoites, by interfering in the functional activity of subtilisin 1 [26]. It is conceivable that similar mechanisms are also important for the invasion process of N. caninum tachyzoites. Moreover, hypothetical proteins without annotated putative functions were identified, and six of the 20 most abundant proteins appear to be involved in transcription and/or translation (Table 3). Table 3. List of the twenty most abundant proteins from Neospora caninum tachyzoites binding to both BKI-1748 and quinine columns, as identified by differential affinity chromatography followed by mass spectrometry. See Supplementary Table S1 for the full dataset. The relative abundance (rAbu) is based on iBAQ, so that the sum or rAbu is 1,000,000 for every column. The proteins are listed according to their decreasing rAbu values in BKI-1748 eluates. In eluates from D. rerio embryos, 98 proteins binding to both columns were identified (Supplementary Table S3). Within this subset, the most abundant protein was the vitellogenin 1 homolog encoded by Q8JH37_DANRE, followed by a lysyl endopeptidase and by another vitellogenin 1 homolog. Besides proteins with diverse functions, six of the 20 most abundant proteins were involved in transcription and/or translation (Table 4). Table 4. List of the 20 most abundant proteins from Danio rerio embryos binding to both BKI-1748 and quinine columns, as identified by differential affinity chromatography followed by mass spectrometry. See Supplementary Table S1 for the full dataset. The relative abundance (rAbu) is based on iBAQ, so that the sum or rAbu is 1,000,000 for every column. The proteins are listed according to their decreasing rAbu values in BKI-1748 eluates. When looking at the specific BKI-1748 binding proteins with proteins binding to BKI-1748 as well as to quinine and evaluating their potential functions in both N. caninum and D. rerio, it was evident that most drug-binding proteins were related to RNA-binding and modification, in particular to RNA splicing, namely 20 of 117 proteins for N. caninum and 37 of 111 proteins for D. rerio. The second most prominent group consisted of proteins binding to and/or modifying other proteins such as peptidases or chaperonins involved in protein refolding with 17 proteins in each group (Table 5). Table 5. Summary of putative functions of BKI-1748 and quinine-binding proteins in Neospora caninum and Danio rerio. The table comprises proteins only identified in BKI-1748 column eluates (see Tables 1 and 2), as well as proteins identified in both BKI-1748 and quinine eluates (see Supplementary  Table S3 for complete list). The functions were identified based on information provided by Uniprot (www.uniprot.org, accessed on 10 November 2021) and related databases.

Discussion
When interpreting differential affinity chromatography data, care must be taken not to underestimate the number of proteins binding to a peculiar compound. If we look at CDPK1, the target enzyme for which BKI-1748 has been created, we see that this protein was identified not only in eluates from BKI-1748, but also from quinine and mock columns. This may be explained by interactions with proteins that bind to the mock columns in an unspecific manner. In a co-immunoprecipitation study using an anti-CDPK1 antibody, we have identified NcMIC3 as a major interaction partner of NcCDPK1, indicating that this kinase may form complexes with other proteins (unpublished data). This protein is found in all eluates, including the mock column (Supplementary Table S1). The number of BKI-binding proteins may therefore be higher than the ones we have seen in our eluates. In all likelihood, the proteins specifically identified in BKI-1748 eluates interacted with the non-quinoline moiety of this molecule only, whereas the proteins identified in both BKI-1748 and quinine eluates may have bound to the quinoline moiety.
Considering the proteins binding specifically to BKI-1748, we identified organism specific subsets, such as hypothetical proteins or alveolin 1 in N. caninum, and proteins specifically involved in development, such as chemotaxin 2 or vitellogenin in D. rerio.
Alveolin 1 is a member of the alveolin family, a class of proteins unique to the phylum alveolata. These proteins are intermediate filament-like proteins and are associated to alveolar sacs located beneath the plasma membrane forming the inner membrane complex, which interacts with the actin-myosin system playing pivotal roles in cell division [27] and motility [28]. As cellular motility in N. caninum is closely linked to host cell invasion, the binding of BKI-1748 to this structural element could contribute to the pronounced inhibition of invasion seen during BKI-treatments. Intracellular tachyzoites that are exposed to BKI-1748 form multinuclear complexes that contain numerous newly formed zoites, which cannot complete the last steps in cytokinesis, thus parasites remain trapped within their host cells [19]. Interference with cell division by binding to alveolin could be partially responsible for this effect.
In D. rerio, a protein with similar functions, namely syntenin, is the second most abundant BKI-1748 specific binding protein. Similar to alveolins, this protein is associated with membrane vesicles, in particular, in neuronal cells, and is involved in cell proliferation, excretion and motility [29]. In zebrafish embryo development, the protein is involved in the spreading and the thinning of the blastoderm closure of the blastopore. Syntenin interacts with syndecan heparan sulphate proteoglycans, which act as co-receptors for adhesion molecules and growth factors. In addition, syntenin binds to phosphatidylinositol 4,5-bisphosphate and with the small GTPase ADP-ribosylation factor 6, which regulate the endocytic recycling of syndecan. Earlier studies have proposed a role for syntenin in the autonomous vegetal expansion of the yolk syncytial layer and the rearrangement of the actin cytoskeleton in extra-embryonic tissues, but not in embryonic cell fate determination [30].
The most abundant BKI-1748 binding protein in D. rerio is chemotaxin 2, which is homologous to mouse leukocyte cell-derived chemotaxin 2 (LECT2). LECT2 was initially isolated as a possible chemotactic factor for neutrophils [31] and has subsequently been demonstrated to be involved in various processes such as liver regeneration and natural killer cell development [32], carcinogenesis [33], and anti-infective and anti-inflammatory responses [34]. However, its actual function in zebrafish embryos remains obscure.
Moreover, confirming our initial hypothesis that common targets may exist in both organisms, we have identified BKI-1748 binding proteins involved in various common pathways with predominant subsets of RNA binding and modifying proteins. Interestingly, the most abundant BKI-1748 specific binding protein was the uridine-rich small ribonuclear binding protein U1 (NcsnrpU1), an essential component of the spliceosome [24]. U1 recognizes specific sequence motifs at the intron-exon boundaries. This would allow generating knockdown mutants by inserting recognition sequences at the 3 -ends of exons of genes of interest [35], thereby validating the function of this protein for apicomplexans. Since the mechanisms of RNA splicing are highly conserved amongst eukaryotes, it is not surprising to find two U1 homologs in the BKI-1748 binding protein fraction from D. rerio, namely, E7F071_DANRE and Q6NUT5_DANRE. These homologs were, however, much longer than NcsnrpU1 with only 44 identical and 76 similar amino acid positions. The homolog with the highest identity to NcsnrpU1, Q4KMD3, is not found in the eluates of any column (Supplementary Table S1). Moreover, the small ribonucleoprotein Sm is another example found in both binding protein subsets. A list of BKI-1748 binding proteins involved in RNA splicing is shown in Table 6. This list may not be complete since in the case of N. caninum, many proteins are not annotated. However, it underlines the impact BKI-1748 (and perhaps compounds with similar structures) could potentially have on RNA splicing. Since RNA splicing is of major importance in embryonic development, interference with this pathway, via mutations of the components or via interaction with chemotherapeutics, may explain the pleiotropic phenotypes observed in D. rerio development, such as the eye and thorax malformations we have observed, and the impact on fertility in pregnant mouse models.
Together, target-based drug development is not a highway to safe drugs without side effects. Off-target effects have to be evaluated for any novel compound using appropriate models. The identification of drug-binding proteins via DAC-as exemplified in the present study-may constitute a rapid way to gain insights into the complex interactions that take place. While it is important to keep in mind that the binding of a drug to a given protein does not automatically imply functional inhibition, DAC still has considerable potential to eliminate undesired side effects by optimizing compounds for the function of their affinoproteomes.

Tissue Culture Media, Biochemicals and Compounds
Cell culture media were purchased from Gibco-BRL (Zürich, Switzerland), and biochemical agents were procured from Sigma St. Louis, MO, USA).

In Vitro Culture and Collection of Parasites
For affinity chromatography necessitating high yields, Neospora caninum NcSpain-7 tachyzoites were grown in HFF as previously described [43,44]. The pellets of tachyzoites were stored at −80 • C until processing.

Maintenance and Collection of Zebrafish Embryos
Zebrafish embryos were kindly provided by the group of Nadia Mercader, from the Institute of Anatomy, University of Bern. Fertilized zebrafish embryos were maintained at 28 • C in 50 mL falcon tubes containing E3 medium (
Fertilized zebrafish embryos were maintained for 48 h allowing the embryos to emerge (change of medium after 24 h) as described above. Subsequently, they were exposed to either 10, 20 or 40 µM of BKI-1748, or were left untreated, during another 48 h. At 96 h post-fertilization, embryos were euthanized as above, and fixed and processed for SEM as described [19].
All specimens were placed onto glass coverslips, sputter-coated with gold and inspected on a Zeiss Gemini450 SEM operating at 5 kV.

Protein Extraction and Affinity Chromatography
For protein extraction, frozen pellets of N. caninum Sp7 tachyzoites and zebrafish embryos were resuspended in ice-cold extraction buffer, i.e., PBS containing 1% Triton X-100 and 1% of Halt proteinase inhibitor cocktail (ThermoFisher). Suspensions were vortexed thoroughly and centrifuged (13,000× g rpm, 10 min, 4 • C). Extraction of pellets was repeated twice. Three ml of extraction buffer was used in total. Supernatants were combined (approximately 3 mg of total protein) and subjected to affinity chromatography.
In order to produce the sepharose matrices conjugated to BKI-1748 or quinine, 0.5 g lyophilized epoxy-sepharose with a C12 spacer was suspended in 15 mL H 2 O and centrifuged at 300× g for 5 min. Washes in water were repeated twice followed by a wash with coupling buffer (0.1 M NaHCO 3 , pH 9.5). After the last wash, 20 mg of each compound dissolved in 2.5 mL DMSO were added and coupling buffer was added to a maximum volume of 5 mL. Mock column medium was generated by incubating 0.5 g of epoxy-sepharose only with DMSO. The mixture was incubated for 3 days at 37 • C under slow but continuous shaking in order to allow coupling of the epoxy group to the compounds. The resulting column medium (approximately 2 mL) was washed with coupling buffer (15 mL) followed by a wash with ethanolamine (1 M, pH 9.5) and by an incubation in 10 mL ethanolamine for 4 h at 20 • C in the absence of light in order to block residual reactive groups. The column medium was then transferred to a chromatography column (Novagen, Merck, Darmstadt, Germany), and extensively washed with PBS-DMSO (1:1) and PBS in order to remove unbound compounds. The columns were stored in PBS containing 0.02% NaN 3 at 4 • C.
Prior to affinity chromatography, mock columns were combined to either quinine or BKI-1748 columns in tandem (mock first, then compound) and washed with 50 mL PBS equilibrated at 20 • C. Crude extracts (3 mL), prepared as described above, were loaded with a flow rate of 0.25 mL/min. The column was washed with PBS until the baseline was flat (10 column volumes, corresponding to ca. 25 mL). Then, the columns were separated, and the binding proteins were eluted with 50 mM acetic acid (5 mL per column). The eluates were lyophilized and stored at −80 • C [45].

Proteomic Analysis of the Eluted Proteins by Mass Spectrometry
The lyophilized eluates were dissolved in 10 µL of 8 M urea and 0.1 M of Tris-HCl -(pH = 8), then, 1 µL of 0.1 M of Tris-Cl -(pH = 8) buffer containing 0.1 M of dithiothreitol were added, followed by incubation for 30 min at 37 • C and constant mixing with 600 rpm. This step was repeated with 1 µL of 0.5 M of iodoacetamide. Iodoacetamide was quenched by the addition of 5 µL 0.1 M of Tris-HCl -(pH = 8) and the urea concentration further diluted to 4 M by the addition of 2 mM calcium dichloride in 20 mM Tris buffer. Proteins were digested for 2 h at 37 • C by the addition of 1 µL of 0.1 µg/µL LysC sequencing grade protease (Promega), followed by further dilution of urea to 1.6 M with above calcium dichloride buffer and 1 µL of 0.1 µg/µL trypsin sequencing grade (Promega). Digestion was completed by incubation over night at ambient room temperature. Digestion was stopped with 2.5 µL of 20% (v/v) trifluoroacetic acid. After an incubation for 15 min at room temperature, the digest was spun for 1 min at 16,000× g, and the cleared supernatant was transferred to a HPLC vial for subsequent nano-liquid reverse phase chromatography coupled to tandem mass spectrometry, as described earlier [46].
The mass spectrometry data were processed with MaxQuant (v1.6.14.0), as described earlier, against a current protein sequence database release from toxodb.org for Neospora caninum (ToxoDB-54_NcaninumLIV_AnnotatedProteins.fasta), and against Uniprot (www. uniprot.org, accessed on 9 December 2021) for Danio rerio. The MaxQuant search results were then further processed in R-studio. Imputation was not applied. Consequently, the output indicated only on-off proteins and labeled those proteins that appeared to be the most extreme outliers in a linear regression comparison. Alignments were performed using the ClustalO tool provided by the Expasy network (www.expasy.org, accessed on 9 December 2021). Modeling of the structure of a binding protein was performed using the Swiss Model homology modeling tools [47] of the Swiss Model repository [48], as provided by the Expasy network.