Identification of substrates for the conserved prolyl hydroxylase Ofd1 using quantitative proteomics in fission yeast

Prolyl hydroxylation functions in diverse cellular pathways, such as collagen biogenesis, oxygen sensing, and translation termination. Prolyl hydroxylation is catalyzed by 2-oxoglutarate (2-OG) oxygenases. The fission yeast 2-OG oxygenase Ofd1 dihydroxylates the 40S ribosomal protein Rps23 and regulates the hypoxic response by controlling activity and stability of the sterol regulatory element-binding protein Sre1. Multiple substrates have been found for 2-OG oxygenases, yet the only known substrate of Ofd1 and its homologs is Rps23. Here, we report the first fission yeast prolyl hydroxylome and demonstrate that hydroxylation is more prevalent than previously known. Using quantitative mass spectrometry, we identify Rpb10, a shared subunit in RNA polymerase I, II, and III, as a novel Ofd1 substrate. In addition, we discovered six Ofd1 binding partners and 16 additional Ofd1 candidate substrates. Although Ofd1 promotes Sre1 degradation, proteomic analysis revealed that Ofd1 does not broadly regulate protein degradation. Instead, the effect of Ofd1 on the proteome is through negative regulation of Sre1N. Finally, we show that the interaction between Ofd1 and the Sre1 bHLH region is conserved across Sre1 homologs suggesting that Ofd1-dependent regulation of SREBPs may be conserved in other fungi. Collectively, these studies provide a new dataset of post-translational modifications and expand the biological functions for a conserved prolyl hydroxylase.


Introduction 30
Eukaryotes require environmental oxygen for essential metabolic processes. Cells exist in

85
Yeast strains were grown to exponential phase at 30°C in yeast extract plus supplements 86 (225 µg/ml each of histidine, leucine, adenine, lysine and uracil) or in SILAC medium 87 (Edinburgh minimal medium plus 75 µg/ml each of histidine, adenine, leucine and uracil and 30 88 µg/ml of regular or heavy lysine) [29].

90
Yeast two-hybrid screen 91 Pairwise yeast two-hybrid assays were performed following the Clontech Matchmaker™ GAL4 92 Two-Hybrid System 3 user manual, using the yeast two-hybrid S. cerevisiae strain AH109 [30].
162 filtered at a 1% FDR. Dynamic modification was chosen for carbamidomethyl C, oxidation M, 163 deamidation of N and Q, hydroxylation of P, dihydroxylation of P, TMT 6-plex for N-term and 164 K.
165 Data processing 166 The protein area calculated by Proteome Discoverer Software was used to quantify protein 167 abundance. The relative abundance of each protein in each of the 10 TMT channels was 168 calculated as following: 170 The relative abundance of WT5 and ofd1Δ2 samples were higher than the other replicates and 171 therefore these two samples were excluded from quantile normalization and further analysis. The 172 remaining eight samples were quantile normalized using 'preprocessCore' package embedded in 173 R [32]. The normalized relative abundance was saved and used for further analysis. Student's t-174 test was applied to calculate the p-value of protein relative abundance in each sample using R.

175
At the peptide level, the precursor area was used to quantify PSM abundance. The 176 relative abundance of each PSM in each of the 8 channels was calculated as following: 242 288 peptides. One peptide from Tif32 (eIF3a) contained a monohydroxylated proline residue with 289 0.340 heavy to light ratio (wild-type/ofd1Δ cells), indicating that the modification is Ofd1-290 independent (Fig S3). Thus, while the yeast two-hybrid screen indicates that Ofd1 binds to eIF3 291 subunits, we did not find evidence that eIF3 subunits are Ofd1 substrates.

292
In addition to subunits of the eIF3m complex, we identified 1263 proteins in the eIF3m

298
We defined prolyl hydroxylated peptides with a heavy to light (H/L) ratio > 3.5 as Ofd1-299 dependent. Using this criterion, two unique prolyl dihydroxylated peptides from Rps23 300 (IGVEAKQP(OH) 2 NSAIRK and IGVEAKQ(deamidated)P(OH) 2 NSAIRK) had H/L ratios of 301 14.941 and 20.030 respectively, confirming that Rps23 P62 dihydroxylation is Ofd1-dependent 302 and validating the dataset. To expand our search for Ofd1 substrates, we repeated the analysis 303 using a FDR of 0.1 and plotted the prolyl hydroxylated peptides based on their H/L and 304 precursor area ( Fig 2D). Again, we defined Ofd1-dependent hydroxylated peptides as having a 305 H/L ratio > 3.5 ( Fig 2D). After manual inspection of the spectra, we identified 16 novel Ofd1 306 candidate substrates, including 9 monohydroxylated peptides, 4 dihydroxylated peptides, and 3 307 peptides containing one mono and one dihydroxylated proline (Table 1, Fig S2). The prolyl 308 dihydroxylated peptide from Nup124, for instance, had complete y-ion series around the proline 309 residue and was more abundant in WT (heavy) cells, indicating the hydroxylation was Ofd1-310 dependent (Fig 2E). For PSMs with incomplete b-or y-ion series, we made specific assumptions 311 about other residue modifications in order to classify these PSMs as prolyl hydroxylated ( 339 Indeed, in WT cells (heavy) > 99% of the peptides from Rps23 were dihydroxylated, whereas in 340 ofd1Δ cells (light) 92% were non-hydroxylated ( Fig S1C). The non-hydroxylated peptides for 341 Fas1 and Rrp5 did not dramatically increase in ofd1Δ cells, perhaps due to the fact that the 342 hydrophobic non-hydroxylated peptides were easier to detect. However, the non-hydroxylated 343 peptide from Rpb10 increased from 52% in WT cells to 96% in ofd1Δ cells, indicating that 344 monohydroxylation of Rpb10 is Ofd1-dependent ( Fig S1C). Together, these data indicate that 345 Rpb10 is an Ofd1 substrate that is present in mono-hydroxylated and non-hydroxylated forms.
346 Additional experiments are required to quantify accurately the two Rpb10 populations. Analysis of the complete eIF3m prep also revealed Ofd1-independent prolyl 348 hydroxylation (H/L ≤ 3.5). Using a strict FDR of 0.01, we discovered 7 prolyl hydroxylated 349 peptides with complete b-or y-ion series around the modified proline residue (

364
In the eIF3m dataset, we observed that a fraction of prolyl hydroxylated peptides (20.9%) 365 deviated from the center with H/L ratios between 0.330 and 0.660 (Fig 2D). This phenomenon 366 was also observed in the entire collection of peptides identified in the eIF3m prep (Fig S4A), 367 indicating that it was independent of prolyl hydroxylation. When H/L ratios were calculated for 368 eIF3m interacting proteins rather than peptides, the left-shifted population disappeared indicating 369 that the phenomenon was likely due to result of noise at PSM level rather than from biological 370 differences between WT and ofd1Δ cells (Fig S4B). In addition, the half-life of proteins 371 containing peptides from the left-shifted population was identical to that of the entire eIF3m prep 372 (Fig S4C), indicating the shift was not due to differences in protein turnover [42]. 379 each strain were used for quantile normalization and further analysis (Fig S5). Overall, 2862 380 proteins were detected, which is comparable to published fission yeast proteomics results (Fig   381 3A) [42]. We searched for prolyl hydroxylation as described previously using a FDR of 0.01 and 382 identified 38 peptides with prolyl dihydroxylation, 122 peptides with prolyl monohydroxylation, 383 and 3 peptides with both mono and dihydroxylation. These PSMs represented a small fraction of 384 total PSM detected (0.2%) and mapped to 128 proteins. After manual inspection, we identified 385 12 peptides with complete b-or y-ion series around the modified proline residue ( 432 (CNJ02310) in C. neoformans (Fig 4A, 4B). Hms1 and Tye7 are SREBP-N proteins that lack the 433 transmembrane segments, suggesting that Ofd1 can regulate SREBP-N proteins in addition to 434 proteolytic product of SREBP [9]. Each of these organisms contains an Ofd1 homolog with To explore whether Ofd1 regulation of Sre1N exists in other fungal organisms, we tested 504 whether this binding interaction is conserved. Yeast two-hybrid analysis between Ofd1 and the 505 bHLH domains of fungal SREBP or SREBP-N proteins showed that four SREBP and SREBP-N 506 proteins from S. cerevisiae, C. albicans, and C. neoformans interacted with Ofd1, suggesting a 507 conserved regulation of SREBPs by Ofd1 in these species (Fig 4). We recently identified an 508

515
Prolyl hydroxylation plays different roles in cells, ranging from a structural function in 516 collagen biogenesis to regulation of the hypoxic response. In yeast, Rps23 is the only known 517 prolyl hydroxylated protein, and hydroxylation functions to regulate Sre1N and promotes 518 translational fidelity [17,18]. In this study, we performed the first search for prolyl 519 hydroxylation in fission yeast, and we discovered 12 additional prolyl hydroxylated proteins with 520 complete mass spectrometry fragmentation around the hydroxylated proline. Even without an 521 enrichment method, 4.2% of proteins were prolyl hydroxylated in our proteomic search, 522 suggesting that prolyl hydroxylation is more prevalent than previously known. The development 523 of a robust enrichment method is critical to fully reveal the prolyl hydroxylome.

524
The 530 Together, these findings suggest that the ribosome is a common target for protein hydroxylation.
531 Further characterization of these modifications and their corresponding oxygenases will provide 532 new insights into translation and cell proliferation.