Differential protein phosphorylation affects the localisation of two secreted Toxoplasma proteins and is widespread during stage conversion

The intracellular parasite Toxoplasma gondii resides within a membrane bound parasitophorous vacuole (PV) and secretes an array of proteins to establish this replicative niche. It has been shown previously that Toxoplasma both secretes kinases and that numerous proteins are phosphorylated after secretion. Here we assess the role of phosphorylation of SFP1 and the related GRA29, two secreted proteins with unknown function. We show that both proteins form stranded structures in the PV that are independent of the previously described intravacuolar network or actin. GRA29 likely acts as a seed for SFP1 strand formation, and these structures can form independently of other Toxoplasma secreted proteins. We show that an unstructured region at the C-terminus of SFP1 and GRA29 is required for the formation of strands and that mimicking phosphorylation of this domain negatively regulates strand development. When tachyzoites convert to chronic stage bradyzoites, both proteins show a dispersed localisation throughout the cyst matrix. Many secreted proteins are reported to dynamically redistribute as the cyst forms and secreted kinases are known to play a role in cyst formation. Using quantitative phosphoproteome and proteome analysis comparing tachyzoite and early bradyzoite stages, we reveal widespread differential phosphorylation of secreted proteins. These data support a model in which secreted kinases and phosphatases are important to dynamically regulate parasite secreted proteins during stage conversion. IMPORTANCE Toxoplasma gondii is a common parasite that infects up to one third of the human population. Initially the parasite grows rapidly, infecting and destroying cells of the host, but subsequently switches to a slow-growing form and establishes chronic infection. In both stages the parasite lives within a membrane bound vacuole within the host cell, but in the chronic stage a durable cyst wall is synthesized that provides protection to the parasite during transmission to a new host. Toxoplasma secretes proteins into the vacuole to build its replicative niche and previous studies identified many of these proteins as phosphorylated. We investigate two secreted proteins and show that phosphorylation plays an important role in their regulation. We also observed widespread phosphorylation of secreted proteins when parasites convert from acute to chronic stages, providing new insight into how the cyst wall may be dynamically regulated.


ABSTRACT 23
The intracellular parasite Toxoplasma gondii resides within a membrane bound 24 parasitophorous vacuole (PV) and secretes an array of proteins to establish this replicative 25 niche. It has been shown previously that Toxoplasma both secretes kinases and that numerous 26 proteins are phosphorylated after secretion. Here we assess the role of phosphorylation of 27 SFP1 and the related GRA29, two secreted proteins with unknown function. We show that both 28 proteins form stranded structures in the PV that are independent of the previously described 29 intravacuolar network or actin. GRA29 likely acts as a seed for SFP1 strand formation, and these 30 structures can form independently of other Toxoplasma secreted proteins. We show that an 31 unstructured region at the C-terminus of SFP1 and GRA29 is required for the formation of 32 strands and that mimicking phosphorylation of this domain negatively regulates strand 33 development. When tachyzoites convert to chronic stage bradyzoites, both proteins show a 34 dispersed localisation throughout the cyst matrix. Many secreted proteins are reported to 35 dynamically redistribute as the cyst forms and secreted kinases are known to play a role in cyst 36 formation. Using quantitative phosphoproteome and proteome analysis comparing tachyzoite 37 and early bradyzoite stages, we reveal widespread differential phosphorylation of secreted 38 proteins. These data support a model in which secreted kinases and phosphatases are 39 important to dynamically regulate parasite secreted proteins during stage conversion. 40 IMPORTANCE

41
Toxoplasma gondii is a common parasite that infects up to one third of the human population. 42 Initially the parasite grows rapidly, infecting and destroying cells of the host, but subsequently 43 switches to a slow-growing form and establishes chronic infection. In both stages the parasite 44 lives within a membrane bound vacuole within the host cell, but in the chronic stage a durable 45 INTRODUCTION

52
The apicomplexan parasite Toxoplasma gondii is widespread infecting approximately one third 53 of the world's population (1). Although Toxoplasma infection is predominantly asymptomatic 54 in healthy hosts, complications occur in the immunocompromised, such as cancer and HIV 55 patients, and in pregnant women. In these patients, infection can cause encephalitis, or 56 severely damage the unborn foetus during pregnancy (2). Additionally, Toxoplasma is a leading 57 cause of retinocharditis (3), and an expansion of strain diversity in South America is causing 58 substantial ocular disease (4). 59 Toxoplasma actively invades nucleated cells of virtually any warm-blooded animal and 60 subsequently replicates within a membrane-bound parasitophorous vacuole (PV). During the 61 acute stage of infection, the rapidly growing tachyzoites complete rounds of invasion and lysis 62 of host cells and disseminate around the body. Upon immune pressure or other stress, the 63 parasite converts to slow-growing bradyzoites that form cysts, predominantly in skeletal 64 muscle and the brain (5). The cyst is a modified PV with a protective cyst wall that allows the 65 parasite to establish chronic, perhaps lifelong infection of the host. 66 expand the analysis to compare the phosphoproteome of acute and chronic stage parasites. 90 We show that many GRAs are differentially phosphorylated between these stages, suggesting 91 that phosphorylation of secreted proteins may be a key determinant for dynamic restructuring 92 of the replicative niche of Toxoplasma. 93

RESULTS 94
A secreted protein forms strand-like structures in the parasitophorous vacuole independently 95 of the intravacuolar network and actin. 96 To investigate the role of phosphorylation of secreted proteins we selected TGGT1_289540 as 97 an ideal candidate. It is substantially phosphorylated after secretion (19) and contains a 98 localised cluster of phosphorylation sites in its C-terminus, allowing for targeted genetic 99 mutagenesis. TGGT1_289540 contains a signal peptide, four coiled-coil domains and an 100 unstructured C-terminus which contains 8 phosphorylation sites (Fig. 1A). Three further 101 phosphorylation sites are located within the rest of the protein. To first verify that 102 TGGT1_289540 is a secreted protein, we expressed a myc-tagged version of the protein which 103 localised to the PV. Interestingly, in contrast to most GRAs that are either found to fill the space 104 between parasites, or are found associated with the PV membrane, TGGT1_289540 appeared 105 in strand-or filament-like structures (Fig.1B). Western-blotting confirmed the presence of a 106 single isoform of myc-tagged protein at the predicted size of ~100kDa in transgenic parasites, 107 but not in a WT control (Fig. 1C). 108 To verify this unusual distribution of TGGT1_289540 in the PV, we generated polyclonal 109 antibodies against the recombinant protein. The antibody recognised a band in Western blot 110 that was absent in WB and IFA when the TGGT1_289540 locus was disrupted (Fig. 1D Examples of Intravacuolar network tubules are indicated by arrows. 128 was the case with the myc-tagged isoform, analysis of the endogenous protein showed a 129 filamentous distribution in the PV (Fig. 1D). 130 As the Toxoplasma PV contains a distinct IVN of membranous tubules we tested whether the 131 strands were related to these structures. The formation of the IVN is dependent on GRA2 so 132 we assessed TGGT1_289540 localisation in RHDgra2 parasites that lack the IVN (14). No 133 obvious difference in filamentous TGGT1_289540 localisation was observed (Fig. 1F), 134 indicating that the filaments are distinct from, and their formation does not depend on the 135 IVN. We therefore named the protein Strand Forming Protein 1 (SFP1 hereafter). 136 We hypothesised that SFP1 strands could function in transport between the parasites and the 137 host cell. Toxoplasma endocytoses proteins from the host cell through an unknown mechanism 138 (16) so we assessed whether SFP1 plays a role in the uptake of proteins from the host cell into 139 the parasite at 3h post infection. As endocytosed material is rapidly digested, this can only be 140 observed upon inhibition of cathepsin protease L function with the inhibitor morpholinurea-141 leucine-homophenylalanine-vinyl phenyl sulphone (LHVS) (16). Under these conditions there 142 was no difference in the proportion of parasites that had taken up Venus fluorescent protein 143 from the host cell indicating SFP1 does not contribute to protein uptake (Fig. S1A). Toxoplasma 144 upregulates host c-Myc in infected cells, a phenotype that is dependent on protein 145 translocation from the PV into the host cell (10,22). To test whether SFP1 plays a role in protein 146 export into the host cell we compared the ability of wild type and Dsfp1 parasites to upregulate 147 host c-Myc. No differences where observed in host cell c-Myc upregulation, indicating that 148 SFP1 does not play a role in protein export from the PV (Fig. S1B). 149 It was recently shown that long filaments of actin extend into the PV during Toxoplasma 150 infection (23). We therefore tested whether treatment with cytochalasin D, to depolymerise 151 actin, would disrupt SFP1 strand formation. We still saw substantial filamentation of SFP1 in 152 the presence of cytochalasin D, indicating that actin is not important in this process (Fig. 1G). 153 While the presence of the IVN is not important for SFP1 filament formation, the opposite could 154 be true, that is, that SFP1 may be important for the formation of the IVN. However, IVN 155 formation was not affected in the absence of SFP1 as shown by transmission electron 156 microscopy ( Fig. 1H). No obvious abnormalities in the overall PV organisation or PVM structure 157 were observed in RHDsfp1 parasites. 158 Collectively these data indicate that SFP1, a protein that was reported to be phosphorylated 159 after secretion, localises to the PV, and forms novel, IVN-and actin-independent filamentous 160 structures. These, as suggested by our data, do not contribute to protein uptake from, or 161 protein export into the host cell. 162 A second SFP1 related protein, GRA29, forms related structures 163 As secreted protein families can form by gene duplication and diversification, we looked for 164 potential paralogs of SFP1 in the Toxoplasma genome using BlastP. We identified a single 165 protein (TGGT1_269690, GRA29) with 24 % identity and a similar architecture ( Fig. 2A & S2) of 166   S3. 182 a signal peptide, coiled-coil domains and, like SFP1, C terminal phosphorylation sites that 183 appear phosphorylated after secretion (19). To compare localisation of this protein with that 184 of SFP1 we generated an HA-tagged version which by IFA localised to the PV and Western blot 185 showed an expected size of ~94 kDa (Fig. 2B,C). In contrast to SFP1 however, we saw smaller 186 filamentous structures and also large puncta associated with the parasites (Fig. 2B) consistent 187 with the localisation reported by Nadiparum et al (24). To investigate whether SFP1 and GRA29 188 were within the same structures, we performed super-resolution microscopy using anti-SFP1 189 antibodies and anti-HA antibodies to localise GRA29. At this improved resolution GRA29 puncta 190 appeared as round spheres or doughnuts, often with SFP1 filaments associated or branching 191 out from them (Fig. 2D). We also observed filamentous SFP1 with smaller dots and strands of 192 GRA29. To further investigate these unusual sphere-like structures we performed correlative 193 light-electron microscopy (CLEM) to identify the subcellular structures. This revealed that 194 GRA29::HA accumulates in electron dense particles in the PV that appear to contain no 195 membrane and resembles proteinaceous aggregates ( Figure S3). 196 To further investigate and verify GRA29::HA structures, we raised polyclonal antibodies against 197 recombinant GRA29. With the antibodies against endogenous protein we observed only short 198 strands and small puncta in the PV that did not resemble the HA-tagged parasite line (Fig. 2E,F). 199 The specificity of the antibodies was verified by Western blot and IFA, which showed the 200 expected molecular weight and positive IFA signal in WT, but not in GRA29 KO parasites (Fig.  201 2F,G). This indicates that the addition of the C-terminal HA tag disrupts GRA29 localisation, 202 inducing protein aggregates, suggesting a potentially important role of the C-terminal region. 203 In order to colocalise endogenous non-tagged SFP1 and GRA29 we generated antibodies 204 against SFP1 peptides in mice which showed the expected reactivity in IFA and Western blot 205 with GRA29 often present at either end of SFP1 filaments (Fig. 2I). 207 Collectively these data show that SFP1 and GRA29 form novel filamentous structures in the PV. 208 C-terminal tagging of GRA29 alters its localisation, either causing or stabilising unusual 209 structures that SFP1 can still associate with. 210 GRA29 initiates SFP1 strand formation, even in the absence of other Toxoplasma proteins 211 As SFP1 and GRA29 appeared to localise within the same structures we hypothesised that they 212 may show an interdependence. To address this, we localised SFP1 in RHDgra29, and GRA29 in 213 RHDsfp1 parasite lines using the specific antibodies raised. Whereas no alteration in GRA29 214 was observed in RHDsfp1, SFP1 formed fewer and longer filaments in the absence of GRA29 215 Phalloidin (magenta) to visualise F-actin, and DAPI (blue). c) HFFs co-expressing Flag-SFP1 and 229 HA-GRA29 show SFP1 strands radiating from GRA29 foci. Immunofluorescence of co-230 transfected HFFs expressing Flag-SFP1 and HA-GRA29. Anti-SFP1 (green), anti-HA (magenta) 231 and DAPI (blue). 232 These data indicate that both SFP1 and GRA29 can form filament-like structures in the absence 233 of other Toxoplasma proteins. Although we were unable to confirm the interaction in co-234 immunoprecipitation or biochemical experiments (data not shown) the co-localisation of SFP1 235 and GRA29 even in this non-physiological context suggests that they can directly interact. 236 Furthermore, the localisation pattern observed in HFFs and the extended SFP1 filaments in 237 parasites lacking GRA29 suggest that GRA29 initiates SFP1 strand formation. 238 The SFP1/GRA29 C terminal tail is required for strand formation 239 The change of localisation of GRA29 upon C-terminal HA-tagging and the cluster of 240 phosphorylation sites in the C-termini of SFP1 and GRA29 suggested the C-terminal tail may 241 play an important role in the regulation of strand-formation. To address this, we expressed 242 truncated SFP1 and GRA29 lacking the unstructured C-terminal region in HFFs. In contrast to 243 forming filaments under these conditions, both proteins appeared distributed throughout the 244 HFF cytoplasm, with sparse foci of aggregation observed (Fig. 4A). This strongly suggested that 245 the C-terminus of the protein plays an important role in filament formation. Consistently, 246 during Toxoplasma infection with RHDsfp1+SFP1DCt there was a lack of typical SFP1 strand 247 formation in the PV (Fig. 4B). However, SFP1DCt appeared aggregated within the PV, rather 248 than dispersed, suggesting other PV resident proteins impact SFP1 distribution. 249 Having established that the C-terminus of SFP1 is important for its subcellular organisation, we 250 asked whether phosphorylation of SFP1 is important for filament formation. To investigate this, 251 we first turned to expression of phosphomutants in HFFs, in which all phosphorylation sites in 252 the SFP1 C-terminus were mutated to Alanine (SFP1_ALA) and phosphomimetics, in which all 253 phosphorylation sites were mutated to Glutamic acid (SFP1_GLU). While SFP-ALA mutants 254 displayed normal filaments in the HFFs, mimicking phosphorylation lead to an inability of SFP1 255 to form filaments. This indicates that phosphorylation may be a negative regulator of filament 256 formation (Fig. 4C). This was replicated when SFP1_ALA and SFP1_GLU were expressed in 257 termini. Anti-Flag or HA antibodies (green), Phalloidin (magenta) to visualise F-actin, and DAPI 263 (blue). b) SIM analysis of HFFs infected with RHDsfp1 expressing SFP1-myc or SFP1DCterm-264 myc. Anti-SFP1 antibodies (green) and DAPI (blue) c) SFP1 phosphomutants (preventing 265 phosphorylation) does not disrupt strand formation. Immunofluorescence analysis of HFFs 266 expressing Flag-SFP1, or Flag-SFP1 with the C-terminal phosphosites mutated to ALA or GLU. 267 Anti-Flag antibodies (green) and DAPI (blue). d) Immunofluorescence analysis of HFFs infected 268 with RHDsfp1 expressing myc tagged SFP1, SFP1 ALA or SFP1 GLU. Anti-myc antibodies (green) 269 and DAPI (blue). 270  If phosphorylation was preventing filament formation, we hypothesized that SFP1 and GRA29 271 expressed in HFFs should not be phosphorylated. To verify that assumption we 272 immunoprecipitated SFP1 from HFFs and analysed the proteins by mass-spectrometry. Despite 273 the presence of many non-modified peptides from SFP1, we observed no phosphopeptides 274 (data not shown), indicating that indeed, SFP1 is not phosphorylated by human kinases when 275 expressed in HFFs. 276 Collectively these data show that the C-termini of SFP1 and GRA29 are important for their 277 regulation and that phosphorylation appears as a negative regulator with the non-278 phosphorylated protein forming strands. 279 Secreted proteins are differentially phosphorylated in the chronic stage 280 To determine whether SFP1 and GRA29 are required in vivo, we generated gene KOs in the 281 Type II strain Pru, and verified gene disruption by Western blot (Fig. S3A). Mice infected with 282 PruDsfp1 and PruDgra29 parasites succumbed to infection similarly to WT PruDku80 parasites 283 indicating neither protein is essential for parasite survival in vivo (Fig. S4B, C). During our 284 experiments with the Pru strains, we observed that in some vacuoles SPF1 and GRA29 285 appeared distributed throughout the PV space instead of forming characteristic strands (Fig.  286 5A). Type II strains differ from the type I RH strain in that they more frequently convert to slow-287 growing bradyzoites in tissue culture. Bradyzoite cysts can be distinguished from tachyzoite 288 vacuoles using fluorescently labelled lectin (DBA) that binds to the cyst wall (25). Parasites with 289 a disperse SFP1/ GRA29 pattern also stained positive for the cyst-wall marker (Fig. 5A). 290 Spontaneous conversion to bradyzoites occurs at low levels in standard tissue culture but high 291 levels of conversion can be induced by pH stress (26). Under bradyzoite-inducing conditions 292 parasite vacuoles were positive for the cyst wall marker CST1, and showed disperse SFP1 and 293 GRA29 indicating that both proteins undergo a major change in localisation between the 294 tachyzoite and the bradyzoite stages (Fig.5B). anti-CST1 (magenta) to visualise the cyst wall, and DAPI (blue). 301

302
We hypothesised that increased phosphorylation of SFP1 and GRA29 under bradyzoite 303 conditions could lead to their dispersion. This phenomenon of redistribution in the bradyzoite 304 cyst has been described for other GRAs and so we wanted to determine if these were also 305 differentially phosphorylated. We therefore used comparative phosphoproteomics to 306 determine differences in the phosphorylation of SFP1 and GRA29 and more broadly between 307 acute (tachyzoite) and chronic (bradyzoite) conditions. Triplicate samples of Toxoplasma were 308 grown in tachyzoite conditions for 27 h or bradyzoite conditions for 3 days (to give comparable 309 numbers of parasites per vacuole) (Fig. 6A). We performed quantitative mass spectrometry to 310 compare protein and phosphosite abundances between the two conditions using tandem mass 311 tags (TMT) as described in (27,28). First, we analysed differential protein abundance and 312 identified 171 differentially regulated proteins with 103 proteins with higher abundance in 313 bradyzoites and 68 with higher abundance in tachyzoites. As expected, known bradyzoite 314 markers such as BAG1, ENO1, LDH2 and MAG1 were more abundant in the bradyzoite samples 315 compared to the tachyzoite samples (Table S1 and  Toxoplasma proteins. For 3730 of the phosphorylation sites we also obtained quantitative 319 proteome data to which phosphorylation site changes were normalized. This allowed us to 320 assess true differential phosphorylation rather than changes originating from differential 321 protein abundance between conditions. After this normalisation step, we identified 337 322 phosphorylation sites that were significantly different between tachyzoite (144 sites more 323 phosphorylated) and bradyzoite (193 sites more phosphorylated) stages (Fig. 6B, Table S1). 324 The 337 phosphorylation sites are found on 170 proteins, 51 of which are predicted to be 325 secreted (ROPs or GRAs in LOPIT dataset, ToxoDB). While GRAs made up 11% of proteins with 326 non-changing phosphosites, they represented 26% of differentially phosphorylated proteins 327 (Fig. 6C). This enrichment of GRAs in the subset of proteins that are differentially 328 phosphorylated between tachyzoite and bradyzoite conditions indicates that these are indeed 329 subject to extensive differential phosphorylation during stage conversion. 330 In this experiment we also identified phosphosites corresponding to the C-termini of SFP1 and 343 GRA29, but they did not significantly change between conditions, (although a previously 344 undetected N-terminal site on GRA29 was detected as more phosphorylated in bradyzoites). 345 However, the GRAs that were detected as differentially phosphorylated included the IVN 346 localised GRA2, GRA4, GRA6, GRA12, and PV membrane GRA1 and GRA5 (Table 1). We also 347 detected differential phosphorylation of the cyst wall proteins CST1, CST3, CST4 and CST6. 348 Here we have analysed the localisation of two secreted phosphoproteins, SFP1 and GRA29 that 366 re-localise during stage conversion to bradyzoites. Both proteins form stranded structures in 367 the tachyzoite PV, that are distinct from the IVN and actin filaments previously observed. We 368 did not identify a function of SFP1 and GRA29 during the lytic cycle or in vivo infections, 369 preventing us from assessing the impact of their phosphorylation in Toxoplasma biology. It 370 could be that the proteins have a subtle effect on growth that we have not been able to 371 measure, or that they are important in another host-species, parasite stage or genetic 372 background we have not tested here. However, neither SFP1 nor GRA29 showed a reduction 373 in fitness in vivo (30) in our recent more sensitive in vivo CRISPR screen, supporting a non-374 essential role under these conditions. 375 We show the importance of the C-termini of both proteins for strand formation. Using 376 mutational analysis, we show that SFP1 phosphorylation is not required for filament formation 377 and that mimicking phosphorylation disrupts filament formation. As such, mimicking 378 phosphorylation mirrored the phenotype observed with removal of the C-terminal tail. This 379 suggests that SFP1 and GRA29 form oligomers dependent on their un-phosphorylated C-380 termini. Phosphorylation negatively regulates the intrinsic behaviour of SFP1 and GRA29 and 381 abrogates their ability to form multimers. However, as both proteins where identified to be 382 phosphorylated after secretion into the PV (19) where, as we show here, they form fibril-like 383 structures, we assume that their phosphorylation is a dynamic process, potentially leading to 384 the regulation of their length and location. It could also be that phosphorylated and non-385 Recently, the PV localised kinase WNG1/ROP35 was shown to phosphorylate several GRAs, 392 including GRA6, and to regulate their membrane association (20). As many GRAs form 393 complexes during trafficking (31), WNG1 phosphorylation is hypothesised to release them 394 from association with a chaperone, freeing them to associate with membranes. As such, WNG1 395 dependent phosphorylation of GRAs was shown to promote their normal behaviour and 396 localisation. We now propose that in addition to WNG1 phosphorylation promoting release 397 and membrane association of GRAs, phosphorylation of secreted proteins can negatively 398 regulate their behaviour. In this way, the parasite could use phosphorylation by kinases within 399 the PV lumen to regulate PV resident proteins. This would allow the parasite to tailor its niche 400 and respond to differing conditions or requirements. 401 Several secreted kinases localise within the PV lumen and are upregulated during the chronic 402 stages of infection (for example ROP21, ROP27, ROP28, and indeed WNG1/ROP35). 403 Importantly, secreted kinases have been shown to contribute to cyst formation in mice 404 indicating that phosphorylation within the PV/forming cyst is required for cyst development. 405 ROP21/27/28 contribute to the parasite establishing chronic infection but their function and 406 targets remain unknown (32). Additionally, a triple knockout of ROP38/29/19 showed a severe 407 defect in cyst formation (33). Here we expand our knowledge of cyst formation by reporting 408 the differential phosphorylation of proteins within the forming cyst. Our data also identifies 409 phosphorylation that is reduced upon stage conversion. The expression of many ROP family 410 kinases is reduced in the chronic stage of infection, which could account for some changes, but 411 it is likely that secreted phosphatases are contributing to the differential phosphorylation. The To generate pSFP1, the TGGT1_289540 promoter and coding sequence were amplified using 458 primers 1 &2 and inserted in to HindIII/PacI digested pGRA (12). A C terminal myc tag was 459 inserted by inverse PCR using primers 3 and 4. pSFP1DCterm was generated by inverse PCR on

MS Data Processing and Analysis 669
The data were searched against Toxoplasma gondii and Homo sapiens (both Uniprot) 670 databases using the Andromeda search engine. Raw data were processed with MaxQuant 671 (version 1.5.2.8). Cysteine carbamidomethylation was selected as a fixed modification. 672 Methionine oxidation, acetylation of protein N-terminus and phosphorylation (S, T, Y) were 673 selected as variable modifications. The enzyme specificity was set to trypsin with a maximum 674 of 2 missed cleavages. The datasets were filtered on posterior error probability to achieve a 675 1% false discovery rate on protein, peptide and site level. Data were further analysed with 676 Perseus (version 1.5.0.9). The data were filtered to remove common contaminants and IDs 677 originating from reverse decoy sequences and only identified by site. Desalted and vacuum dried samples were solubilized in 1 ml of loading buffer (80% acetonitrile, 713 5% TFA, 1 M glycolic acid) and mixed with 5 mg of TiO2 beads (Titansphere, 5 µm GL Sciences 714 Japan). Samples were incubated for 10 min with agitation, followed by a 1 min 2000 × g spin 715 to pellet the beads. The supernatant was removed and used for a second round of enrichment 716 as explained below. Beads were washed with 150 μl loading buffer followed by two additional 717 washes, the first with 150 μl 80% acetonitrile, 1% TFA and the second with 150 μl 10% 718 acetonitrile, 0.2% TFA. After each wash, beads were pelleted by centrifugation (1 min at 2000 719 × g) and the supernatant discarded. Beads were dried in a vacuum centrifuge for 30 minutes 720 followed by two elution steps at high pH. For the first elution step, beads were mixed with 100 721 μl of 1% ammonium hydroxide (v/v) and for the second elution step with 100 µl of 5% 722 ammonium hydroxide (v/v). Each time beads were incubated for 10 min with agitation, and 723