A positive feedback loop mediates crosstalk between calcium, cyclic nucleotide and lipid signalling in Toxoplasma gondii

Fundamental processes of obligate intracellular parasites, such as Toxoplasma gondii and Plasmodium falciparum, are controlled by a set of plant-like calcium dependent kinases (CDPKs), the conserved cAMP- and cGMP-dependent protein kinases (PKA and PKG), secondary messengers and lipid signalling. While some major components of the signalling networks have been identified, how these are connected remains largely unknown. Here, we compare the phospho-signalling networks during Toxoplasma egress from its host cell by artificially raising cGMP or calcium levels to activate PKG or CDPKs, respectively. We show that both these inducers trigger near identical signalling pathways and provide evidence for a positive feedback loop involving CDPK3. We measure phospho- and lipid signalling in parasites treated with the Ca2+ ionophore A23187 in a sub-minute timecourse and show CDPK3-dependent regulation of diacylglycerol levels and increased phosphorylation of four phosphodiesterases (PDEs), suggesting their function in the feedback loop. Disruption of CDPK3 leads to elevated cAMP levels and inhibition of PKA signalling rescues the egress defect of ΔCDPK3 parasites treated with A23187. Biochemical analysis of the four PDEs identifies PDE2 as the only cAMP-specific PDE among these candidates, while the other PDEs are cGMP specific, two of which are inhibited by the predicted PDE inhibitor BIPPO. Conditional deletion of the four PDEs supports an important, but non-essential role for PDE1 and PDE2 in growth, with PDE2 controlling A23187-mediated egress. In summary we uncover a positive feedback loop that enhances signalling during egress and links several signalling pathways together.


39
The Apicomplexa are obligate intracellular parasites that pose a considerable challenge  Irrespective of the egress trigger, it is clear that secondary messengers play a key role in 56 driving the process forward once initiated. Indeed, calcium (Ca 2+ ) (Carruthers and Sibley,

136
Generation of Calcium Reporter Lines to align BIPPO and A23187 signalling 137 pathways. 138 To investigate how the cGMP and calcium signalling pathways converge and differ, we 139 compared their phosphorylation dynamics using two activators of these pathways: 140 BIPPO, a PDE inhibitor, and the calcium ionophore A23187. 141 The signalling kinetics following Ca 2+ ionophore and BIPPO treatment vary, so we first 142 determined a timepoint at which both pathways should be comparable. Common to both 143 treatments is a raise in intracellular calcium levels before egress. We therefore chose 144 peak intracellular calcium levels as a reference point to facilitate a direct comparison 145 between BIPPO-and A23187-treated parasites. To this end, we generated a stable 146 calcium sensor line that co-expresses, through use of a T2A ribosomal skip peptide, an 147 internal GFP control and the genetically encoded ruby Ca 2+ biosensor jRCaMP1b (Alves 148 et al., 2021) from a single promoter (Fig. 1A-B). The expression of the biosensor did not 149 have any discernible effects on Ca 2+ ionophore (A23187) or BIPPO induced egress rates immobilised parasites illustrated distinct Ca 2+ response curves; BIPPO treatment led to a 155 rapid increase in Ca 2+ levels, (Fig. 1Ci), while the cytosolic Ca 2+ rise detected upon 156 A23187 treatment appeared more gradual (Fig. 1Cii). To facilitate optimal alignment, and to account for the rapid kinetics of these signalling pathways, treatment timings of 15s 158 (BIPPO) and 50s (A23187) were selected for subsequent phosphoproteomics 159 experiments.

161
Phosphoproteomic Responses at temporally aligned calcium flux. 162 Having identified the optimal BIPPO and A23187 treatment timings to achieve maximal 163 calcium release, we wanted to identify and compare phosphorylation events that take 164 place during BIPPO-and A23187-induced signalling cascades at these timepoints. We line. This allowed us to identify ΔCDPK3 dependency of signalling events during A23187 175 and BIPPO induced egress and is explained further below. 176 We quantified changes in phosphorylation states by calculating the log2-transformed 177 intensity ratios (log2FC) of A23187-or BIPPO-stimulated WT parasites versus DMSO-178 treated WT parasites (DataS1). In total we quantified 7,811 phosphorylation sites across 179 these conditions.  The rapid signalling progression upon treatment with BIPPO and A23187 inevitably 185 results in variability in phosphosite intensities between replicates, where despite our best 186 efforts, signalling may be stopped with several seconds difference between experiments. 187 As such variability results in poor p-values in classical t-tests and, by extension, an under- of mean Ca 2+ response (jRCaMP1b/GFP normalised to 0), following addition of (i) 50µM BIPPO or (ii) 8 µM phosphoproteomic experiments. Data was collected from ≥ 10 vacuoles (in separate wells) over ≥6 days.

197
ΔCDPK3 parasites treated with 8 µM A23187 (50s) were included to facilitate later analyses. (E) Correlation 208 samples are shown. In (i) Coloured data points highlight sites that were differentially up-or down-regulated 209 (red and blue, respectively) upon BIPPO, but not A23187 treatment, while in (ii) coloured data points 210 highlight sites that were differentially up-or down-regulated (red and blue, respectively) upon A23187, but 211 not BIPPO treatment. Orange dotted lines represent 3xMAD outlier thresholds used to determine differential 212 site regulation (log2FC>0.5 for up-regulated sites and log2FC<-0.5 for down-regulated sites).  reporting of true treatment-regulated sites, we did not subject DR sites to further p-value-219 based thresholding. However, the reporter intensities associated with DR sites correlated 220 well across replicates (r>0.89, Supp Fig. 3B). This suggests that despite some of the 221 aforementioned replicate variability, the overall trends across replicates were consistent, 222 and these scores could therefore be confidently averaged to provide values that are 223 representative of a site's phosphorylation state at the timepoint of interest.

224
Comparison of the log2FCs observed in BIPPO-and A23187-treated samples shows 225 strong correlation between the phosphorylation responses of these conditions (r=0.9039) 226 (Fig. 1E), suggesting that the signalling pathways at these selected timepoints align 227 sufficiently well to directly compare them.

229
Induced Signalling. 230 To investigate the signalling events that are shared between or are unique to BIPPO and 231 A23187 treatment, we identified DR sites for each treatment condition. We then identified 232 DR sites that were successfully quantified in both treatments, which allowed us to 233 examine their behaviour under both conditions. In total we identified 746 BIPPO and 981 234 A23187 DR sites. A large overlap was detected between treatments for both up-and 235 down-regulated phosphosites (DR UP and DR DOWN , respectively); ~91% of phosphosites 236 up-regulated following BIPPO treatment showed similar regulation upon A23187 addition 237 and ~58% of BIPPO down-regulated sites behaved similarly following A23187 treatment 238 (Fig. 1F). 239 We also observed some dissimilar regulation between conditions; 59 phosphorylation 240 sites were found to be up-regulated following BIPPO treatment only, while 237 sites were 241 phosphorylated exclusively following A23187 treatment. Of the DR DOWN phosphorylation 242 sites, 22 were found to be unique to BIPPO treatment, while 79 were unique to A23187 243 treatment. These treatment-specific sites may originate from distinct signalling pathways, 244 activated by each of the compounds. To discern whether these disparate site behaviours 245 are truly treatment-specific effects, or whether they are the result of imperfect alignment 246 of the treatment timings, we visualised phosphorylation site log2FCs following A23187 247 treatment, and highlighted phosphorylation sites that were only DR following BIPPO 248 treatment (Fig. 1Gi). Similarly, we also visualised phosphorylation site log2FCs following 249 BIPPO treatment, and highlighted phosphorylation sites that were only DR following 250 A23187 treatment (Fig. 1Gii). In both instances, most sites approached the DR thresholds 251 for up-or down-regulation. While this does not preclude the possibility that some of the 252 BIPPO-and A23187-specific DR UP/DOWN sites are regulated in a drug-exclusive manner, 253 it is likely that the majority of these sites would pass the DR threshold within seconds, 254 and that minor changes in treatment timing can make the difference between surpassing 255 the DR threshold or not. 256 Collectively, these findings demonstrate that at temporally aligned calcium release within 257 the parasite, it is nearly impossible to detect clear signalling features that confidently 258 distinguish the BIPPO-activated signalling pathway from the signalling cascade activated 259 upon cytosolic Ca 2+ elevation by A23187 treatment.

261
Signalling Pathway. 262 A substantial overlap between BIPPO-regulated and A23187-regulated sites was 263 expected given the increase of cytosolic Ca 2+ in both treatment conditions. However, the 264 inability to confidently distinguish BIPPO from A23187 signalling was unexpected, as 265 these agents are believed to initiate egress by activating distinct, albeit interconnected,  (Fig. 1H).

278
The differential phosphorylation of several proteins in the upstream pathway of cNMP 279 production/regulation hints at a putative Ca 2+ -mediated feedback loop that regulates 280 cGMP and/or cAMP signalling. The existence of such a feedback mechanism could 281 account for our inability to confidently discern PKG-specific signalling upon BIPPO 282 treatment, as such signalling would be activated upon treatment with both BIPPO and 283 A23187.

293
To identify phosphorylation sites that might fit such criteria, we wanted to identify 294 phosphorylation sites that are CDPK3-dependent upon A23187s treatment, but CDPK3-295 independent upon BIPPO treatment. To do this, we generated a ∆CDPK3 parasite line by . As expected, we found that A23187-induced egress was substantially inhibited 302 in this line (Fig. 2Bi), while a less severe egress delay was observed in BIPPO-treated 303 ∆CDPK3 parasites (Fig. 2Bii). This recapitulates previous findings (Lourido,Tang and 304 David Sibley, 2012) that activation of PKG partially compensates for a loss of CDPK3. 305 We next quantified phosphorylation events in ΔCDPK3 parasites treated with DMSO, 50 306 µM BIPPO (15s) or 8 µM A23187 (50s) at 37°C (Fig. 2C) in biological replicates (2x 307 DMSO, 3x A23187, 3x BIPPO). In this experiment (set 2), we included 2 biological 308 replicates of BIPPO-treated WT parasites. In conjunction with the ionophore-treated 309 ΔCDPK3 parasites included in set 1 (Fig. 1D), this allowed us to identify CDPK3-310 dependent phosphorylation sites during BIPPO-and A23187-induced egress (Data S2). 311 We first identified DR phosphorylation sites across all datasets for which we had  The 16 phosphorylation sites that show CDPK3 dependency exclusively upon A23187 318 treatment constituted putative candidates for PKG/CDPK3 substrate overlap. We 319 reasoned that, if a DR UP site was found to be CDPK3-dependent upon A23187 treatment 320 only, this phosphorylation site should be recovered in ∆CDPK3 parasites following BIPPO 321 treatment. Only 3 of these phosphorylation sites showed this behaviour and were located 322 on two hypothetical proteins (TGGT1_243460, TGGT1_232120) and a DnaJ domain-323 containing protein (TGGT1_203380) (Data S2). While these findings do not completely 324 rule out the 'substrate overlap' theory to account for BIPPO's compensatory effects, the 325 putative overlap is extremely small, and none of these proteins contain predicted domains 326 that would explain the rescue of CDPK3 mutants by BIPPO in induced egress.

327
Collectively, our current findings provide more evidence for a Ca 2+ -regulated feedback 328 loop model than for PKG/CDPK3 substrate overlap.

359
DR thresholds were set at 3x MAD of the log2FC across each WT timepoint (15s, 30s 360 and 60s). Phosphorylation sites were considered differentially regulated if at any given 361 timepoint their log2FC surpassed these thresholds. CDPK3 dependency was determined 362 for each phosphorylation site by calculating the log2 ratios of A23187-treated WT and 363 ∆CDPK3 parasites for each timepoint. The resulting ratios were used to calculate the 364 MAD at each timepoint, and the most stringent score was used to set 3X MAD outlier 365 thresholds. A DR site was considered CDPK3-dependent if, at any given timepoint, it 366 simultaneously passed the DR and CDPK3 dependency thresholds. 367 We identified 2,408 phosphorylation sites (DR UP ) upon A23187 treatment in WT 368 parasites, which were also quantified in ∆CDPK3 parasites (Data S3). To examine 369 whether this dataset recapitulates our previous findings, we compared the DR UP sites   Table S1 for all

407
In the up-regulated clusters, we identified a preponderance for phosphorylation motifs 408 with arginine in the -3 position (Fig. 3Di), a consensus sequence that has previously been 409 shown to be preferentially phosphorylated by CDPK1 (Lourido et al., 2013) and possibly 410 CDPK3 (Treeck et al., 2014). Reassuringly, this consensus motif was also observed 411 among DR UP CDPK3-dependent sites (Fig. 3Dii). Down-regulated phosphorylated sites, 412 meanwhile, show a clear enrichment for proline in the +1 position ( Fig. 3Diii & Fig. 3Div). 413 This indicates that while CDPK activity (and/or activity of kinases with a similar substrate 414 preference) is being induced by calcium-signalling, a distinct set of one or more kinases 415 with this phosphorylation motif preference is being inactivated concurrently. Alternatively, 416 this could be mediated by the activation of a specific phosphatase. We observed that to ionophore, but also the dynamics of phosphorylation are similar between them. 434 We also found significant enrichment of membrane proteins in CDPK3-dependent   It is important to note that visualisation of the CDPK3-dependent DR UP clusters revealed 459 that for many sites, the effect of CDPK3 deletion was temporary, such that there was an 460 initial delay in phosphorylation, but by 60s the sites reached only slightly lower log2FC 461 values than in WT (Fig. 3C). This may point to the redundant or compensatory activity of 462 a protein kinase other than PKG, which could in part account for the fact that egress still 463 occurs in CDPK3-depleted parasites, albeit at a delayed pace. Such a delay was not 464 clearly detectable in the CDPK3-dependent DR DOWN clusters.

465
The GO terms "signal transduction" and "hydrolase activity" mentioned above contained or 60 seconds before quenching to stop the signal chain, followed by lipid analysis. A 0s 498 (DMSO) control was also included. After 60 seconds of A23187 stimulus, WT parasites 499 produced slightly more DAG than ∆CDPK3 parasites, however there was no difference 500 with the DMSO control (Fig. 4A, Supp Fig. 6A). Accordingly, WT parasites began to show 501 less phospholipids than ∆CDPK3 parasites after 45 seconds of stimulus (Fig. 4A, Supp 502 Fig. 6B), consistent with a lack of turnover of phospholipids to produce DAG in the 503 ∆CDPK3 knockout parasites. While DAG-related proteins were the primary lipid related 504 proteins affected in ΔCDPK3 based on our timecourse phosphoproteome, we also 505 identified other proteins involved in palmitoylation and triacylglycerol synthesis that were 506 differentially regulated, so we further investigated other lipids including Free Fatty Acids 507 (FFAs) and triacylglycerols (TAGs). We observed a trend towards an increase in the 508 levels of free fatty acids (FFAs) in WT parasites following A23187 stimulus (Fig. 4B, Supp 509 Fig. 6C), which remained unchanged in ΔCDPK3 parasites. This was accompanied by a 510 concomitant change in triacylglycerols (TAGs) whereby prior to stimulus, ΔCDPK3 511 parasites had more TAGs than WT parasites, but after A23187 stimulus, WT tachyzoites 512 produced more TAGs over time so that levels became similar between both parasite lines 513 (Fig. 4C, Supp Fig. 6D).This shows that following A23187 treatment, ∆CDPK3 parasites 514 have altered FFA and TAG abundance necessary for lipid recycling and storage, 515 consistent with a speculated role for CDPK3 in metabolic regulation (Treeck et al., 2014).  (Fig. 4D).

532
In contrast, and somewhat surprisingly, we found no changes in cGMP levels following 533 A23187 treatment over the course of 60 seconds. It is important to note that our    Basal cAMP levels, by contrast, were 8.3% higher in knockout parasites compared to WT.

561
Following treatment with A23187, we found that cAMP levels initially rose in both WT and 562 knockout parasites at 5 seconds post treatment, with a gradual decrease to below basal 563 levels at 60 seconds post treatment (Fig. 4E). We observed no immediate change in  Intriguingly, when we investigated A23187-induced egress rates of the remaining 577 intracellular H89 pre-treated parasites, we found that the egress delay normally observed 578 in ∆CDPK3 parasites was largely rescued with H89 pre-treatment (Fig. 4Gi-ii). This finding 579 suggests that pharmacological inhibition of cAMP signalling is sufficient to partially 580 compensate for the deletion of CDPK3. Cell biological and biochemical characterisation of PDE1, 2, 7 and 9. 585 The preponderance of A23187-induced phosphorylation on several PDEs suggests that 586 they may play an important role in the cAMP-and cGMP-mediated signalling cascades 587 that lead to egress assuming that phosphorylation may directly, or indirectly control their 588 activity. Specifically, we predicted a cAMP-specific PDE would play an important role 589 given the rescue of the egress defect observed in ΔCDPK3 parasites by the PKA inhibitor 590 H89. As the specificity for the majority of Toxoplasma PDEs has not been experimentally 591 validated, we generated HA-tagged conditional knockouts (cKOs) of the 4 PDEs identified 592 as being phosphorylated following A23187 treatment in order to characterise them and 593 identify which are capable of hydrolysing cAMP (Fig. 5A). For each line, integration of 594 both repair templates was validated by PCR (Supp Fig. 7). Western blot analysis 595 confirmed that they migrate at their predicted sizes (Fig. 5B), and we found that each PDE 596 occupies a distinct cellular localisation (Fig. 5C), in agreement with a previous report (Vo to be dual-specific as a positive control (Flueck et al., 2019). PDEs 1, 7 and 9 were able 601 to hydrolyse cGMP, while PDE2 is specific for cAMP (Fig. 5D). Only PfPDEβ displayed 602 dual-hydrolytic activity in our hands.

603
To further confirm the hydrolytic specificity of the PDEs, we treated each of the samples 604 with BIPPO. While the specificity of BIPPO has not been experimentally validated in 605 Toxoplasma, cGMP-specific PDE1 and 7 were significantly inhibited by BIPPO, while the 606 cAMP-specific PDE2 was refractory to BIPPO inhibition (Fig. 5E). Interestingly, PDE9, a 607 cGMP-specific PDE appears less sensitive to BIPPO treatment. This is in agreement with 608 a previous study (Vo et al., 2020), although in our hands PDE9 is cGMP-specific and not 609 dual-specific. Collectively these data show that BIPPO is a cGMP-specific inhibitor in 610 Toxoplasma and lends further support that PDE2 is a cAMP-specific PDE. Interestingly, 611 we found that both the cAMP-and cGMP-hydrolysis activities of PfPDEβ are inhibited 612 with BIPPO. It will be interesting in the future to evaluate the structural differences 613 between the PDEs and the inhibitory potential of BIPPO.

638
Functional assessment of the candidate PDEs reveals that PDEs 1 & 2 are 639 important but not essential during the lytic cycle while PDEs 7 & 9 are dispensable. 640 We next wanted to establish which of the aforementioned PDEs were essential for lytic 641 growth. Addition of rapamycin (RAP) to the HA-tagged PDE cKO lines leads to excision 642 of the PDE gene of interest in the respective cKO lines (Fig. 6A). Despite near complete 643 excision in all lines as observed by PCR 24 hours post RAP treatment (Fig. 6B), it was 644 only until 3 days post-treatment that we saw complete protein depletion below detectable 645 levels ( Fig. 6C-D). Therefore, all subsequent experiments were conducted with parasites   parasites. Using a medium-throughput plate-based egress assay we found that only 685 deletion of PDE2 showed a modest egress defect (Fig. 6H, Supp Fig. 8). This suggests 686 an important, but non-essential role for PDE2 in cAMP signalling. Since the egress defect 687 observed upon deletion of PDE2 did not reach the severity of the egress defect observed 688 in ΔCDPK3 parasites, it is likely that other PDEs, and/or cyclases are involved in the 689 dysregulation of cAMP levels found in our study, some of which may have been missed 690 in our phosphoproteome or may not be regulated by phosphorylation. To test this, we 691 generated a ΔCDPK3 KO in the PDE2 cKO line using a similar approach to the one used 692 in Fig. 2A, however we substituted the HXGPRT cassette with a DHFR-TS selection 693 cassette. Upon deletion of PDE2, we observed a further increase of cAMP levels 694 compared to either single gene deletion (Fig. 6I) suggesting that CDPK3 is regulating 695 another unknown cAMP-specific PDE, or an adenylyl cyclase.

697
In this study we have aimed to unravel the complexity of the signalling pathways that 698 govern the control of host cell egress of Toxoplasma from its host cell. Several signalling 699 components conserved in higher eukaryotes have previously been identified, and their 700 connectivity, to some extent described. However, the published data is not currently is possible that our collective inability to observe elevated cGMP levels is explained by 726 the sensitivity limits of the assay employed, it is similarly possible that cAMP-mediated 727 signalling is exerting its effects on the PKG signalling pathway in a cGMP-independent 728 manner. While no such mechanism has been described, it is possible that 729 phosphorylation of PKG may lead to changes in its affinity for cGMP or it may regulate 730 the activity of the kinase itself. Further work is needed to clarify the role of cGMP levels 731 in these conditions. We also identified dysregulation of DAG and phospholipid signalling 732 in ΔCDPK3 parasites following A23187 treatment, which could be contributing to the 733 delayed egress phenotype observed in the KO parasites. Having identified CDPK3-734 dependent phosphorylation sites on both DGK1 and PI-PLC in our timecourse 735 phosphoproteome, it is possible that these perturbations are being directly mediated by 736 CDPK3. Alternatively, and as outlined above, any changes in cGMP levels or PKG activity 737 in ΔCDPK3 parasites could also lead to the dysregulation of phospholipid signalling we 738 observed. 739 We identify PDE2 as one contributor of cAMP control, however, through double gene 740 deletions of PDE2 and CDPK3, we show that other cAMP signalling components likely 741 contribute to a further cAMP imbalance. This could be either via as yet unidentified PDEs 742 with cAMP specificity or, more likely, an adenylate cyclase. In support of the latter, we 743 identified several CDPK3-dependent phosphorylation sites on an ACβ following A23187 744 treatment. We also found that deletion of PDE2 alone leads to a modest egress 745 phenotype that does not reach ΔCDPK3 levels. This is not surprising: our data, and  However, we did not detect any CDPK3-dependent phosphosites on PDE2, so a direct 750 link between CDPK3 and PDE2 is currently missing. However, it is possible that CDPK3-751 dependent phosphorylation sites on PDE2 were not detected in our mass-spectrometry 752 experiments for technical reasons, or that PDE2 is indirectly regulated by CDPK3. It has, 753 for instance, been reported that CDPK3 promotes egress by phosphorylating the egress 754 suppressor SCE1 (McCoy et al., 2017). Deletion of SCE1 in ΔCDPK3 parasites largely 755 rescues several CDPK3-dependent phosphosites, suggesting that another SCE1-756 suppressed kinase is able to partially compensate for loss of CDPK3, and likely amplifies 757 regular CDPK3-mediated egress under normal conditions.

758
The research community is also continuously identifying novel components involved in 759 signalling which, once identified, could shed light on how different pathways are 760 interconnected. For example, a recent report has identified SPARK, a novel kinase that 761 appears to mediate Ca 2+ release in a PKG-dependent manner and can be largely 762 bypassed via treatment with A23187 (Smith et al., 2021). While A23187 treatment 763 appears to restore absolute levels of Ca 2+ release in SPARK depleted parasites, the rate 764 of both calcium release and egress remains partially delayed. These findings suggest that 765 PKG-regulated SPARK still contributes, to some degree, to A23187-mediated egress.

766
This observation is in keeping with our proposition that A23187 signalling feeds back into 767 the PKG signalling pathway.

768
While our study provides strong evidence for a CDPK3-mediated feedback loop to control 769 rapid egress and nearly overlapping signalling pathways at peak calcium flux, we cannot 770 draw conclusions about the signalling events at the onset of calcium release.    Table S6. The calcium sensor construct 799 was generated as recently described (Alves et al., 2021). The construct was linearised 800 using NaeI and transfected into RH ∆ku80∆hxgprt parasites as described previously 801 (Soldati and Boothroyd, 1993) to generate the GFP-T2A-jRCaMP1b calcium sensor line. µg/mL) to culture medium. Integration of the HXGPRT cassette at the CDPK3 locus was 818 confirmed using primer pairs 8/9 and 10/11 to confirm 5' and 3' integration respectively.

819
Absence of the endogenous CDPK3 locus was confirmed using primers 12/13.   imaged across ≥ 7 days for each condition. Image analysis was performed using Nikon 886 NIS-Elements analysis software. jRCaMP1b and GFP signals at 0s were set to 0 (zero) 887 and 1 respectively. jRCaMP1b/GFP was used as a readout for ∆Ca 2+ .