Theileria parasites subvert E2F signaling to stimulate leukocyte proliferation

Intracellular pathogens have evolved intricate mechanisms to subvert host cell signaling pathways and ensure their own propagation. A lineage of the protozoan parasite genus Theileria infects bovine leukocytes and induces their uncontrolled proliferation causing a leukemia-like disease. Given the importance of E2F transcription factors in mammalian cell cycle regulation, we investigated the role of E2F signaling in Theileria-induced host cell proliferation. Using comparative genomics and surface plasmon resonance, we identified parasite-derived peptides that have the sequence-specific ability to increase E2F signaling by binding E2F negative regulator Retinoblastoma-1 (RB). Using these peptides as a tool to probe host E2F signaling, we show that the disruption of RB complexes ex vivo leads to activation of E2F-driven transcription and increased leukocyte proliferation in an infection-dependent manner. This result is consistent with existing models and, together, they support a critical role of E2F signaling for Theileria-induced host cell proliferation, and its potential direct manipulation by one or more parasite proteins.

www.nature.com/scientificreports www.nature.com/scientificreports/ uninfected controls 8 ; however, the functional role of E2F signaling in Theileria-induced host cell proliferation has not yet been thoroughly characterized.
Here, we leveraged a comparative genomics approach to predict parasite protein sequences that are potentially involved in the Theileria-induced transformation of host leukocytes. We discovered a peptide derived from a parasite protein that has the sequence-specific and infection-specific ability to increase host E2F signaling by binding to the Retinoblastoma-1 (RB), a negative regulator of E2F signaling, and thereby contributing to host cell hyper-proliferation. Hence, cumulative evidence to date demonstrates the critical role of the RB/E2F signaling axis in Theileria host cell proliferation, and suggests a mechanism by which one or more parasite proteins could manipulate E2F signaling.

Results
We developed a bioinformatics pipeline to perform a comprehensive prediction of potential host-transforming, short linear motifs in Theileria proteins. This pipeline was based on the rationale that such peptide motifs have all of the following attributes: (i) are in proteins predicted to be exposed to the host (i.e. secreted), (ii) are present in host-transforming Theileria species (T. parva and T. annulata), and (iii) are found in proteins that are expressed in the schizont life-cycle stage based on RNA-seq data. In order to reduce false positives, we also imposed a strict criterion that the protein motif be absent in the protein sequences of orthologs found in other apicomplexan genomes, including non-transforming Theileria (T. orientalis and T. equi). While only 400 protein orthologs were unique to T. parva and T. annulata, a total of 109 protein motifs, present in these 400 proteins, satisfied the required criteria (Fig. 1a,b; Supplementary Dataset). This list is critically dependent on a recently generated T. parva genome annotation, since the coding sequences of 101 of these 400 genes were altered in the new annotation (Tretina et al., submitted) 9 .
Interestingly, we identified 15 T. parva proteins that contain a short linear motif predicted to interact with the well-studied host tumor suppressor Retinoblastoma-1 protein (RB) (Fig. 1c, Supplementary Table 1). Along with p107 and p130, RB is a pocket family protein that plays a critical role in mammalian cell cycle regulation by inhibiting E2F transcription factor activity, an important regulator of cell proliferation and survival 10 . The central RB pocket domain is highly conserved between humans and bovids ( Supplementary Fig. 1) and has two significant binding sites: a) the interface of the A and B cyclin folds, which binds to the E2F transactivation domain LXXLFD motif, and (b) a three-helix cleft of the B cyclin fold, which binds to chromatin-modifying proteins containing an LXCXE motif, such as histone deacetylase 1 (HDAC1) 10 . Out of our 15 predicted T. parva RB-binding proteins, eight had the LXCXE motif, and seven had the LXXLFD motif (Supplementary Table 1).

Figure 1. Comparative genomics identified 15
Theileria parasite proteins with the potential to alter host E2F signaling, one of which can bind Retinoblastoma-1 pocket domain. (a) A heatmap of orthologous proteins that are shared by each pair of apicomplexan parasites used in this study. (b) A heatmap of short eukaryotic linear motifs in secreted proteins that are shared by pairs of apicomplexan parasites used in this study. (c) 15 Theileria secreted proteins have short linear motifs that are present in the Eukaryotic Linear Motif (ELM) database and are predicted to interact with host Retinoblastoma-1 (RB). (d) A kinetic surface plasmon resonance binding curve (sensorgram) showing that peptides derived from one of the proteins in (c) binds RB in a sequencespecific manner, as shown by surface plasmon resonance. (e) Listed are estimated dissociation constants for the LXCXE peptide (same as (d)), a full-length Theileria protein containing that peptide (TpMuguga02_00667; TpGcpE), as well as for host E2F1, binding to the RB pocket domain.
Since the proliferation of T. parva-infected host cells is thought to be critical for the pathogenesis of infection, we investigated whether synthesized peptides, each 9 to 19 amino acids long and containing these 15 T. parva motifs, could bind to RB (Supplementary Table 1). Surface plasmon resonance (SPR) was used to screen these peptides for specific RB1 pocket domain-binding activity. Recombinant, purified RB pocket domain was coupled to the surface of the sensor chip, and purified peptides were introduced at various concentrations to quantify kinetics if response curves indicated a detectable protein-protein interaction. Sequence-scrambled peptides were used as negative controls to exclude false-positive signals. This approach led to the discovery of two peptides that bind RB in a sequence-specific fashion, one with an LXCXE motif (in TpMuguga_02g00667, TpGcpE, Fig. 1d,e), and one with an LXXLFD motif (in TpMuguga_02g02355). SPR kinetics assays revealed that full-length, recombinant TpGcpE protein also binds the RB pocket domain with nanomolar affinity (18.7 nM) (Fig. 1e).
The discovery of these peptides led us to investigate the particular RB/E2F signaling components in Theileria-infected cells. Eight E2F transcription factor genes have been identified in mammals, which can be divided into two groups: activators (E2F-1, E2F-2, E2F-3a), which can bind RB, and repressors (e.g. E2F-3b, E2F-4, E2F-5), which can bind RB, p107 and p130. The balance between the activator and repressor E2Fs seems to be primarily controlled by the phosphorylation status of the associated pocket proteins to which they bind. E2F interactions with the Dimerization Partner (DP) ligands are also important for regulating E2F transcription factor activity 11 . We used Western blot to compare the levels of RB family, E2F, and DP-1 protein between T. parvatransformed cells that were left untreated with those treated with buparvaquone to kill the parasite and induce an arrest in host cell proliferation. By comparing expression levels between the buparvaquone-treated and untreated groups, we can determine parasite-dependent changes in the expression levels of these proteins. We found that the T. parva-transformed B cell line TpMD409.B2 expresses RB, p130, E2F-1, E2F-3, E2F-4, DP-1 (Fig. 2a). It is noteworthy that phosphorylated RB (pRB) and p130 migrate faster in parasite-killed TpMD409. B2 cells than in untreated cells (Fig. 2a). These faster migrating bands likely correspond to hypophosphorylated, proliferation-suppressive forms of RB and p130, at 24 h and 48 h after treatment with buparvaquone. Interestingly, after 48 h of treatment, but not 24 h, E2F-1 and E2F-3 protein levels decrease, a slower-migrating DP-1 band increases, but E2F-4 levels remain unaltered (Fig. 2a). This indicates that, in the presence of live intracellular T. parva parasites, host cells express higher E2F-1 and E2F-3 levels, DP-1 has decreased post-translational modifications, and RB and p130 are hyper-phosphorylated. Furthermore, the binding of E2F to DNA appears to increase in the presence of live parasites, since treatment with the parasiticidal drug buparvaquone over 24 h and 48 h is accompanied by an overall decrease in E2F DNA binding activity, as well as by an increase in complexes migrating at slow and intermediate speed bound to an E2F-dependent promoter, which likely represent E2F/RB repressor complexes (Fig. 2b). Proliferation was also confirmed to be dependent on the parasite by BrdU incorporation of untreated and buparvaquone-treated T. parva infected cells (Fig. 2c), as has been shown previously 8 .
We also found that T. parva-transformed B cells, TpMD409.B2 cells, contain significant levels of E2F transcriptional activity, since cells transfected with an E2F-driven luciferase reporter plasmid (E2F-luc) exhibited a 200% increase in luciferase activity compared to cells transfected with a plasmid containing a mutated E2F binding site (mutated) (Fig. 3d). Since no system for genetic manipulation is available yet for intracellular Theileria, we instead incubated T. parva-and T. annulata-infected bovine leukocytes (transfected with a luciferase reporter regulated from an E2F promoter) with synthesized peptides containing the RB-binding LXCXE motif, or a mutated LXNXE negative control, linked to a cell-penetrating peptide. If the LXCXE-containing peptide can regulate E2F activity in a sequence-specific manner, we would expect it to increase E2F activity, and that this effect would be reduced or abrogated in the mutated LXNXE peptide-treated group. As expected, the LXCXE peptide increased E2F activity in T. parva-infected B cells, as well as in T. annulata-infected B cells and macrophages. In contrast, the mutated LXNXE peptide did not show the same reporter activity as the LXCXE peptide ( Fig. 3b,d,e), providing direct functional evidence that the parasite-derived LXCXE motif can induce E2F activity in Theileria-infected leukocytes.
Since the LXCXE peptide can induce E2F activity, we investigated whether it could, on its own, increase bovine leukocyte proliferation. To extend our observations to T. annulata-infected leukocytes, we incubated T. annulata-infected B cells with WT or mutated LXNXE peptides and counted cell numbers at 24 h and 48 h. The addition of these peptides had no effect on cell viability in any of the experimental groups (data not shown). Importantly, only the LXCXE peptide induced a significant increase in proliferation of T. annulata-infected B cells as well as (to a much lesser extent) of uninfected, but immortalized controls (Fig. 3b). Interestingly, an E2F inhibitor (HLM006474) specifically reduced the proliferation of T. annulata-infected TBL20 cells and not uninfected BL20 cells (Fig. 3c), an effect that was not dependent on cell death (data not shown). These data demonstrate that RB/E2F interactions significantly regulate Theileria-driven host leukocyte proliferation.
In order to map the site of the interaction of the LXCXE motif with the RB pocket domain, we used SPR to test the ability of full-length TpGcpE protein to compete with two proteins that have known, well-characterized interactions with the RB pocket domain, E2F1 and HDAC1. The RB pocket domain was coupled to the surface of the sensor chip, and purified, recombinant, full-length TpGcpE was introduced into the flow channel to bind the RB pocket domain to saturation; then either purified, recombinant E2F-1 or HDAC1 was added to the channel and the response curve recorded to determine if they competed withTpGcpE binding. We found that E2F-1 does not compete with full-length TpGcpE for binding to the RB pocket domain (Fig. 4a,b). However, TpGcpE does compete with HDAC1 for RB (Fig. 4c,d), despite a > 4.5-fold higher concentration of the latter, suggesting that the full-length TpGcpE can compete with endogenous chromatin-modifying enzymes like HDAC1 for binding to the LXCXE-binding cleft of the RB pocket domain. We also found that RB interacts with HDAC1 in uninfected cells, and to a lesser extent, in infected cells, as detected by co-immunoprecipitation (Fig. 4b,e,f). This could be partially explained by competition of HDAC1 with LXCXE-containing proteins, as total RB and HDAC1 levels do not seem to be altered by the presence of T. parva (Fig. 4e). (2020) 10:3982 | https://doi.org/10.1038/s41598-020-60939-x www.nature.com/scientificreports www.nature.com/scientificreports/ Discussion RB binds and represses E2F transcription factor family proteins in non-proliferating cells, in part by recruiting LXCXE motif-containing chromatin modifiers to E2F-dependent promoters. Mitogenic signals lead to an increase in cellular cyclin-dependent kinase activity, which phosphorylates RB and thereby inactivates RB-dependent E2F repression by dissociating the RB-E2F complex. Interestingly, the RB LXCXE-binding cleft is also critical for repressing transcription from the cyclin E and cyclin A promoters. Since cyclin E/A-CDK2 complexes can phosphorylate and inactivate RB, this provides an important positive feedback signal for proliferation 12 . Free E2F proteins form heterodimers with E2F dimerization partner proteins DP-1 or DP-2 to form a functional transcription factor that regulates cell proliferation and survival 10,13 .
While full-length TpGcpE was measured to have very strong binding to host RB by SPR, the LXCXE-containing peptide derived from TpGcpE bound with lower affinity than the full-length protein (Fig. 1). This is indicative of the growing consensus that although the LXCXE motif can be used to identify potential RB-interaction partners, structural studies with both pathogen-derived and mammalian host-derived proteins indicate that there are other important determinants of high-affinity interactions with the pocket cleft 14 . For example, out of the eight LXCXE motif-containing peptides testing in this study, only one bound RB. This is supported by other studies finding that several proteins that interact with this region of RB do not contain an obvious LXCXE motif 15,16 , suggesting that RB interactions in the LXCXE-binding cleft are more complicated than currently understood from structural studies. While TpGcpE was measured to have very strong binding to host RB by SPR (Fig. 1), immunofluorescence imaging with a polyclonal antibody ( Supplementary Fig. 2) found very little signal outside of the parasite co-localizing with host RB. There can be a few alternative, non-competing explanations for our limited ability to detect GcpE in the host cell cytosol: (a) the signal peptide in GcpE is also predicted to contain a motif targeting it to the apicoplast (a chloroplast-like organelle specific of the Apicomplexa), suggesting that only small amounts of GcpE remain in the host cell cytosol, and at levels below antibody detection; (b) during host cell division the multinucleated macroschizont is partitioned into new daughter host cells and enough GcpE gets released into the cytosol to sustain E2F activation; or (c) antagonism of the RB/HDAC1 binding by Theileria parasites results from competition with more than one parasite-derived motif, since we identified 15 T. parva proteins predicted to interact with RB, where eight had the LXCXE motif and seven had the LXXLFD motif. Moreover, SPR demonstrated that the LXCXE motif (in TpMuguga_02g00667, TpGcpE), and the LXXLFD motif (in TpMuguga_02g02355) bound the RB pocket domain with nanomolar affinity. Examining all 15 proteins for their presence in the host cell cytosol and eventual interaction with RB will be necessary to test hypothesis (c). It should also be noted that the parasite marker used for staining the schizont (p104), is secreted by the parasite and then subsequently localizes to the parasite surface, suggesting that this staining may illuminate more than just the parasite intracellular compartment. However, there is some weak GcpE staining outside of the p104-stained regions, indicating that some GcpE may localize to the host nucleus.
Our demonstration that penetrating peptides linked to the LXCXE motif in GcpE can block RB-HDAC1 binding, activate E2F-driven transcription and promote infected leukocyte proliferation, argues that targeting the RB/E2F signaling pathway could lead to novel chemotherapeutics against T. annulata-induced pathogenesis. For example, drugs have recently been developed to prevent binding of LXCXE motif-containing HPV E7 protein to wild-type E2F binding construct), or mutated E2F-luc constructs (labeled 'mutated') together with a CMV-lacZ plasmid to normalize for transfection efficiency. Luciferase and β-galactosidase activities were measured 24 hrs after transfection. E2F-luc relative luciferase activity normalized to mE2F-luc. Shown is the average of 3 independent experiments with standard deviation. BL20 or TBL20 cells were treated with (b) 1μM of penetrating peptides (VKKKKIKREIKIYIEEVFTPLVLKCKELK-K(FITC)) containing the TpGcpE LXCXE motif, or an LXNXE control (VKKKKIKREIKIYIEEVFTPLVLKNKELK-K(FITC)) or (c) E2F inhibitor or DMSO control, and then counted by hemacytometer after the indicated timepoints. Shown is the average of 3 independent experiments with standard deviation. Then, either (d) virulent Ode T. annulata-transformed macrophages, or (e) BL20/TBL20 T. annulata-transformed cells were transfected with E2F-luc plasmid (or mutated E2F-luc as in (c)) and incubated for 24 hrs at 37 °C. Then, cells were treated with penetrating peptides as in (b) for 2 hrs and luciferase activity was quantified. Data shown are representative of three independent experiments done with biological duplicates, with mean + standard deviation (*P-value < 0.05 compared to untreated; ‡ P-value < 0.05 compared to its respective BL20 control).

Scientific RepoRtS |
(2020) 10:3982 | https://doi.org/10.1038/s41598-020-60939-x www.nature.com/scientificreports www.nature.com/scientificreports/ host RB and are selectively cytopathic to HPV-infected cells 17 . In fact, the E2F inhibitor HLM0006474 was able to specifically inhibit the proliferation of T. annulata-infected, but not uninfected bovine B cell lines (Fig. 3c), indicating that E2F signaling could be a viable target of chemotherapeutic intervention for this infection. Since T. parva and T. annulata bind to their host's mitotic spindle and divide in synchrony with the host leukocytes 18 , host proliferation is intricately linked to parasite proliferation. This unique mechanism of parasite proliferation also provides the parasite with immune protection, as it avoids parasite egress from the host cell and exposure to the host immune response. We have established that uncontrolled proliferation of Theileria-transformed leukocytes can be achieved through modulation of RB/E2F signaling, and that TpGcpE has the potential to alter those interactions. Conclusive demonstration that specific Theileria proteins manipulate host RB/E2F signaling in vivo awaits future studies.

Methods
Bioinformatic analyses. Comparative genomics. Jaccard-filtered eukaryotic clusters of orthologous genes (KOGs) were generated as described previously 19 , using the following genomes from EuPathDB (August 25 th , 2014) 20 : Babesia bovis T2Bo, Cryptosporidium hominis TU502, Cryptosporidium muris RN66, Cryptosporidium parvu, Iowa II, Neospora caninnum Liverpool, Plasmodium falciparum 3D7, Plasmodium knowlesi strain H, Plasmodium vivax Sal-1, Theileria annulata strain Ankara, Theileria equi strain WA, and Theileria orientalis strain Shintoku. Also used for creating the KOGs were the updated Theileria parva strain Muguga annotation and an updated Babesia microti RI annotation recently completed by our group 21 . Localization predictions were made with TargetP 22 , transmembrane domain predictions with TMHMM 23 , GPI predictions were made with GPI-SOM 24 , all using the default parameters. The entire ELM database was downloaded from http://elm.eu.org/ and custom scripts were used to search the proteome of each Apicomplexan in the Sybil database using regular expressions in Python and a p-value cutoff of 0.001.  Kinetics analysis of binding. To minimize mass transport effects, the binding analyses were performed at flow rate of 30 µl per minute at 25 °C. The analytes (60 µl each, 0-25 µM in HBS-EP buffer with 0.05% P20) were injected into flow cell and the association of analyte and ligand were recorded respectively by surface plasmon resonance (SPR) with a Biacore T200 (Biacore, Inc., New Jersey). After this, the surface was washed with buffer for 180 seconds to follow the dissociation of analyte-ligand complexes. The signal from the blank channel (flow cell-1) was subtracted from the channel containing human Retinoblastoma-1 protein. The binding was removed by injecting 100 µl of HBS-P, pH 7.4.
Data analysis. Sensorgrams of the interaction generated by the instrument were analyzed using the software BIAeval 3.2 (Biacore Inc., New Jersey). The reference surface data were subtracted from the reaction surface data to eliminate refractive-index changes of the solution, injection noise and non-specific binding to the blank surface. A blank injection with buffer alone was subtracted from the resulting reaction surface data. Data was globally fitted to the steady state affinity.
Cell culture and transient transfections. The TpMD409.B2 cell line is a T. parva Muguga-infected B-cell clone (B2) whose establishment, phenotypic characteristics and culture conditions have been described [25][26][27] . BL20 cells are a retrovirus-transformed B cell line isolated from an infected cow. To eliminate the parasite, cells were treated for the indicated time with buparvaquone (50 ng/ml, 1 mg/ml stock in ethanol; Sigma). To inhibit E2F activity, cells were treated for the indicated time with HLM006474 (10 μM in DMSO, Sigma), or DMSO (10μM; Sigma) as a control. The T. annulata-infected cell lines BL20 (B cell), as well as Ode (macrophage) cell lines are well described and characterized.
Flow cytometric analysis of BrdU incorporation. TpMD409.B2 cells were transfected as described for cell cycle analysis except that H2B-EGFP expression vector (10 μg) was used instead of phGFP-S65T. 48 h after transfection, cells were pulse labeled for 1 h with BrdU (5-bromo-2′-deoxyuridine) (10 μg/ml) then harvested and fixed in 1% paraformaldehyde, 0.01% Tween 20 for 48 h to 72 h at 4 °C. BrdU staining was performed using phycoerythrin-labeled anti-BrdU antibody (Pharmingen) according to the previously published DNase I procedure 29 . A minimum of 2,000 events in each whole population and GFP positive fraction were collected on a FACSCAN flow cytometer and analyzed with CellQuest.
Immunofluorescence assays. 5 × 10 4 cells were plated on glass coverslips coated with poly-L-lysine. Cells were fixed with 4% paraformaldehyde in PBS for 10 min, and permeabilized using 0.5% Triton X-100. After a 30 min treatment with blocking solution (1% BSA and PBST buffer: 0.1% Tween20), cells were stained for 2 h with the primary antibodies (anti-RB: Santa Cruz Biotechnology #sc-73598, anti-1C12, and a rat anti-GcpE antibody. After three washes with PBS, cells were incubated with the secondary antibody (Alexa fluor 488 donkey anti-rat IgG #A21208 and Alexa fluor 594 goat anti-mouse IgG #A11005) for 45 min in the dark. After three additional washes, slides were stained with 4-6-diamidino-2-phenylindole, diluted 1/1000 (Sigma Aldrich) for 5 min to visualize the nuclei. Labelled preparations were mounted in Dako and analysed by inverted microscopy (Leica DMI6000s). Acquisitions were made with the Metamorphous software.