Combining Recombinase-Mediated Cassette Exchange Strategy with Quantitative Proteomic and Phosphoproteomic Analyses to Inspect Intracellular Functions of the Tumor Suppressor Galectin-4 in Colorectal Cancer Cells

Galectin-4 (Gal4) has been suggested to function as a tumor suppressor in colorectal cancer (CRC). In order to systematically explore its function in CRC, we established a CRC cell line where Gal4 expression can be regulated via the doxycycline (dox)-inducible expression of a single copy wildtype LGALS4 transgene generated by recombinase-mediated cassette exchange (RMCE). Using this model and applying in-depth proteomic and phosphoproteomic analyses, we systematically screened for intracellular changes induced by Gal4 expression. Overall, 3083 cellular proteins and 2071 phosphosites were identified and quantified, of which 1603 could be matched and normalized to their protein expression levels. A bioinformatic analysis revealed that most of the regulated proteins and phosphosites can be localized in the nucleus and are categorized as nucleic acid-binding proteins. The top candidates whose expression was modulated by Gal4 are PURB, MAPKAPK3, BTF3 and BCAR1, while the prime candidates with altered phosphorylation included ZBTB7A, FOXK1, PURB and CK2beta. In order to validate the (phospho)proteomic data, we confirmed these candidates by a radiometric metabolic-labelling and immunoprecipitation strategy. All candidates exert functions in the transcriptional or translational control, indicating that Gal4 might be involved in these processes by affecting the expression or activity of these proteins.


Introduction
Galectins are members of a family of 15 carbohydrate-binding proteins that contain one or two carbohydrate-binding domains specific for galactose-containing glycans. Galectins are widely expressed in various cell types and mediate their functions both intra-and extracellularly. They are involved in the regulation of cell growth, differentiation, signal transduction, apoptosis, pre-mRNA splicing, cell-cell and cell-matrix interactions, cellular polarity, motility and innate/adaptive immunity [1,2]. Considering their important physiological functions, it is not surprising that an abnormal expression of galectins has been implicated in several disease states. In particular, galectins have been shown to play a fundamental role in cancer development, progression and metastasis [2][3][4] tract. It belongs to the tandem-repeat subfamily of galectins, which are composed of two identical carbohydrate recognition domains (CRDs) at the N-terminus and C-terminus joined by a short linker sequence (the "hinge" region). The CRDs distinctly differ in the binding preferences for their ligands, which encompass glycosphingolipids, oligosaccharides and glycoproteins bearing lactosyl structures. The physiological functions of Gal4 are not fully resolved, but their roles in the stabilization of lipid rafts, cell adhesion, apical trafficking in enterocytes and intestinal wound healing have already been established [5]. These examples suggest galectin's broad intracellular activity, as well as its functionality at the cell surface. Likewise, the roles of Gal4 in tumor development and progression may also be ascribed to intra-and extracellular actions. For instance, the loss of cell surface binding of Gal4 has been shown to promote colorectal tumorigenesis by affecting the expression of the genes involved in growth control and apoptosis [6,7], while the addition of Gal4 to the culture medium of a set of colorectal cancer cell lines profoundly reduced their proliferation rates and induced differentiation. In these detailed proteomic and phosphoproteomic analyses, broad effects of exogenous Gal4 on the cell's molecular phenotype, due to expression changes in a large number of proteins, are indicated. The bioinformatic interpretation of these results by protein ontology analyses supports the concept of Gal4's action as a tumor suppressor [8]. Moreover, additional phosphoproteomic analyses revealed that exogenous Gal4 has a bearing on malignancy-associated intracellular protein phosphorylation [9].
Besides their presence in the extracellular milieu and on the outer surface of cells, galectins have been detected in relatively high concentrations intracellularly, e.g., in the cytosol or the nucleus [10]. While the binding of galectins to the cell surface is considered to be mediated by binding to membrane-anchored surface receptors via their CRDs, intracellular actions may involve both CRD-dependent and -independent mechanisms [11]. In colorectal, pancreatic and hepatocellular cancer, intracellular Gal4 appears to exert tumor-suppressing activity by the inhibition of protumorigenic signaling pathways [12][13][14]. However, how the loss of Gal4 expression in CRC supports tumorigenesis on the molecular level is still poorly understood.
In order to systematically address this topic, we now established a CRC cell line where Gal4 expression can be regulated via the doxycycline (dox)-inducible expression of a single copy wildtype LGALS4 transgene generated by recombinase-mediated cassette exchange (RMCE). Using this model system and pursuing a quantitative proteomics and phosphoproteomics approach, we sought to identify target proteins whose expression and phosphorylation is regulated specifically by Gal4.

Generation of Colorectal Cancer Model Cell Line with Dox-Inducible Gal4 Expression
In order to investigate the role of the expression loss of Gal4 in CRC, we established a colorectal cancer cell line model system that enables the inducible reconstitution of Gal4 expression in an isogenic background. The model cell line is derived from Gal4-deficient cell line HCT116 by recombinase-mediated cassette exchange (RMCE). The HCT116 HygTK master cell line carries a Hyg-TK expression cassette flanked by two recombination sites (F3/F), which confers hygromycin resistance. Two independent clones of this master cell line have been established in our lab previously, differing in the integration site of the expression cassette, which is either localized on chromosome 1 (C1orf159) for clone #5 and on chromosome 5 (ALDH1L1 gene, silenced by methylation in HCT116 cells) for clone #22 [15,16]. The luciferase-Gal4 expression cassette was inserted directly into these defined loci by recombinase-mediated cassette exchange, resulting in HCT116-Gal4 #5 and #22 cells. In these model cell lines, Gal4 and luciferase expression are regulated by the bidirectional dox-inducible promoter (P tet bi) that is comprised of two minimal promoters, human cytomegalovirus (hCMV) and mouse mammary tumor virus (MMTV)-2. This promoter allows the initiation of gene transcription in the presence of doxycycline (dox) by allowing the binding of the reverse transcriptional transactivator to seven Tet operators within the promoter construct ( Figure 1A, Tet-On system). The established cell line therefore enables the specific investigation of the biological relevance of the loss of Gal4 in CRC tumorigenesis. Two independent clones, showing strong dox-inducible luciferase activity, as well as Gal4 expression on the protein level, were chosen for detailed characterization and further study ( Figure 1B-D).
rectional dox-inducible promoter (Ptetbi) that is comprised of two minimal promoters, human cytomegalovirus (hCMV) and mouse mammary tumor virus (MMTV)-2. This promoter allows the initiation of gene transcription in the presence of doxycycline (dox) by allowing the binding of the reverse transcriptional transactivator to seven Tet operators within the promoter construct ( Figure 1A, Tet-On system). The established cell line therefore enables the specific investigation of the biological relevance of the loss of Gal4 in CRC tumorigenesis. Two independent clones, showing strong dox-inducible luciferase activity, as well as Gal4 expression on the protein level, were chosen for detailed characterization and further study ( Figure 1B-D).

Characterization of Doxycycline-Inducible HCT116-Gal4 Cell Lines
The induction of Gal4 expression in two independent monoclonal cultures of HCT116 #5 and #22 was examined in a dose-and time-dependent manner. Each clone showed a slightly different dose-dependency from Gal4 expression. Clone #22 showed a higher induction in luciferase activity than #5 on the protein level, while clone #5 exhibited a more efficient induction of Gal4 re-expression (Figure 2A,C). For clone #5, the induction of Gal4 expression was on a comparable level in the range of 0.25-1 µg/mL Dox. In all further experiments, a Dox concentration of 0.5 µg/mL was chosen for HCT116 Gal4 #5 cells to ensure a strong Gal4 expression. Clone #22 shows stable and strong Gal4 expression at about 0.75 µg/mL Dox and higher. Consequently, 0.75 µg/mL of Dox was used for all subsequent experiments with this clone. The time-course analysis of both model cell lines showed that Gal4 was detectable after 12 h and reached a peak within 72 h. In the absence of Dox, no Gal4 protein was detectable ( Figure 2B,D). Proliferation assays performed on both model cell line clones, as well as the master cell line HCT116-HygTk and parental HCT116 cells, showed no significant influence of the chosen doxycycline concentrations on cell growth. Similarly, reconstituted Gal4 expression did not have any effect on the cell proliferation rate of these cells ( Figure S1). In order to check for Gal4 secretion to the extracellular space, we examined the presence of Gal4 in conditioned media of both HCT116-Gal4 clones. In addition to the presence of Gal4 in lysates of dox-treated cells, Gal4 was also clearly detectable in and, thus, secreted onto their medium upon dox treatment ( Figure 3), whereas it was completely absent in the medium of the untreated cells.

Characterization of Doxycycline-Inducible HCT116-Gal4 Cell Lines
The induction of Gal4 expression in two independent monoclonal cultures of HCT116 #5 and #22 was examined in a dose-and time-dependent manner. Each clone showed a slightly different dose-dependency from Gal4 expression. Clone #22 showed a higher induction in luciferase activity than #5 on the protein level, while clone #5 exhibited a more efficient induction of Gal4 re-expression (Figure 2A,C). For clone #5, the induction of Gal4 expression was on a comparable level in the range of 0.25-1 μg/mL Dox. In all further experiments, a Dox concentration of 0.5 μg/mL was chosen for HCT116 Gal4 #5 cells to ensure a strong Gal4 expression. Clone #22 shows stable and strong Gal4 expression at about 0.75 μg/mL Dox and higher. Consequently, 0.75 μg/mL of Dox was used for all subsequent experiments with this clone. The time-course analysis of both model cell lines showed that Gal4 was detectable after 12 h and reached a peak within 72 h. In the absence of Dox, no Gal4 protein was detectable ( Figure 2B,D). Proliferation assays performed on both model cell line clones, as well as the master cell line HCT116-HygTk and parental HCT116 cells, showed no significant influence of the chosen doxycycline concentrations on cell growth. Similarly, reconstituted Gal4 expression did not have any effect on the cell proliferation rate of these cells ( Figure S1). In order to check for Gal4 secretion to the extracellular space, we examined the presence of Gal4 in conditioned media of both HCT116-Gal4 clones. In addition to the presence of Gal4 in lysates of dox-treated cells, Gal4 was also clearly detectable in and, thus, secreted onto their medium upon dox treatment ( Figure 3), whereas it was completely absent in the medium of the untreated cells.   all subsequent experiments with this clone. The time-course analysis lines showed that Gal4 was detectable after 12 h and reached a peak w absence of Dox, no Gal4 protein was detectable ( Figure 2B,D). Prolife formed on both model cell line clones, as well as the master cell line HC parental HCT116 cells, showed no significant influence of the chosen do trations on cell growth. Similarly, reconstituted Gal4 expression did n on the cell proliferation rate of these cells ( Figure S1). In order to check to the extracellular space, we examined the presence of Gal4 in conditio HCT116-Gal4 clones. In addition to the presence of Gal4 in lysates of Gal4 was also clearly detectable in and, thus, secreted onto their mediu ment (Figure 3), whereas it was completely absent in the medium of th

Gal4 Induces Proteomic and Phosphoproteomic Changes
As both clones of our model cell line HCT116 Gal4 showed similar growth and expression characteristics, only one clone (HCT115-Gal4 #5) was chosen for the subsequent detailed analyses of Gal4-induced proteomic and phosphoproteomic changes. For this purpose, the SILAC (stable isotope-labeling with amino acids in the cell culture) approach for mass spectrometry-based proteomics and phosphoproteomics was applied. These experiments were conducted in three biological replicates. HCT116 Gal4 #5 cells were labeled with "heavy medium" (containing L-[ 13 C 6 , 15 N 4 ] arginine (R10); L-[ 13 C 6 , 15 N 2 ] lysine) or "light medium" (containing L-[ 12 C 6 , 14 N 4 ] arginine (R0); L-[12C6,14N2] lysine). After exhaustive labeling for 8 days, the 'heavy' cells were treated with doxycycline for 72 h for induction of Gal4 expression, while 'light' cells served as the untreated control ( Figure S2A). A total of 3083 proteins were identified and quantified in at least two biological replicates (Table S2 and Figure S2B). Twenty proteins were regulated (fold change > 1.5) in a Gal4-dependent manner (Table 1). In parallel, 2071 phosphorylation sites were identified with a high probability (>0.75) and quantified in at least two biological replicates. To each phosphosite, corresponding information from the PhosphoSite Plus database was aligned (Table S3). In order to avoid the false identification of phosphorylation changes due to alterations in whole protein expression, 1603 phosphosites were matched and normalized to their protein expression levels (Table S4). Among them, 48 phosphosites, located on 21 phosphoproteins, were found to be regulated (>1.5-fold change) in at least two biological replicates (Table 2).  When the regulated proteins and phosphoproteins were mapped together and visualized by the STRING database in an interaction network extracted from the entire human genome ( Figure 4A), the resulting network consisted of 540 nodes and 9808 edges. A Gene Ontology enrichment analysis showed strong enrichment in proteins involved in RNA binding (p = 9.23 × 10 −126 ) and nucleic acid binding (p = 1.29 × 10 −111 ). The enrichment analysis also revealed that most of the regulated proteins and phosphoproteins can be localized in the nucleoplasm (p = 1.24 × 10 −127 ), nuclear lumen (p = 7.16 × 10 −127 ) and nucleus (p = 5.26 × 10 −114 ).
In order to check the validity of the data obtained by quantitative (phospho)proteomics, we selected four representative candidates, each from the proteomic and phosphoproteomic analysis for validation by an independent method. To get a sensitive and reliable quantitation of the candidates, we combined radiometric metabolic labeling with immunoprecipitation. A dual metabolic labeling strategy was applied to account for potential alterations in protein expression on the data obtained with a phosphoproteomic analysis. To this end, 35 S-methionine was used to label the protein part, while 32     the observed protein count in each category and calculated p-value corrected for multiple testing according to Benjamini and Hochberg. Likewise, the altered phosphorylation of ZBTB7A (Zinc Finger and BTB Domain Containing 7A), FOXK1 (Forkhead Box K1), PURB and CK2beta (Casein Kinase II subunit beta) could be corroborated ( Figure 6). Interestingly, PURB was upregulated at the protein expression level, accompanied by a decreased phosphorylation level. Likewise, the altered phosphorylation of ZBTB7A (Zinc Finger and BTB Domain Containing 7A), FOXK1 (Forkhead Box K1), PURB and CK2beta (Casein Kinase II subunit beta) could be corroborated ( Figure 6). Interestingly, PURB was upregulated at the protein expression level, accompanied by a decreased phosphorylation level.

Discussion
In previous studies, the transient transfection of LGALS4 into pancreatic carcinoma [13], gastric cancer cells [17] oligodendrocytes [18] and CRC cell lines [7] was already used to investigate the intracellular functions of Gal4. In order to overcome the disadvantages of these hitherto applied transfection strategies, we, for the first time, applied a RCME-based strategy and inducible target gene expression to test Gal4 functions in a cancer cell line, which brings some benefits. In particular, only a single copy of the LGALS4 transgene is integrated into the genome of the model cell line. The expression of the transgene can be regulated by dox, thereby allowing the induction of the physiological levels of Gal4 in an isogenic background. The careful characterization of two independent HCT116-Gal4 clones proofed the inducibility of Gal4 expression in a time-and dose-dependent manner. Thus, the engineered HCT116-Gal4 cell lines provide a versatile tool to analyze functional consequences and (patho)biological relevance of altered Gal4 expression in CRC.
The induction of intracellular Gal4 expression did not affect the growth rate of the cells. By contrast, Satelli et al. [7] described a reduced proliferation after an overexpression of Gal4 in HCT116 cells. This might be explained by the different gene doses of our model system versus the transient transfection strategy. Moderate amounts of Gal4 were detectable in the conditioned medium of HCT116Gal4 #5 and #22, indicating that the galectin is secreted by the model cell lines. Since the secretion of galectins is mediated independently from the classical secretory pathway by a not yet fully defined nonclassical pathway [19,20], our cell lines may become a versatile tool to address this in future studies. The binding of Gal4 to the extracellular surface of CRC cells has been shown to exert a strong inhibitory effect on proliferation [8]. However, the secretion rates in our model system obviously do not generate extracellular concentrations that are sufficient to affect cell growth.
In order to systematically screen for intracellular changes induced by Gal4 expression, we quantified the proteomic and phosphoproteomic changes in our model system. Broad coverage of the cellular proteome was achieved by quantitating more than 3000 cellular proteins. In parallel, over 2000 phosphorylation sites were identified and quantified. Combining proteomic and phosphoproteomic analyses brought up the advantage that alterations in protein expression could be taken into account in the protein phosphorylation analysis. A bioinformatic evaluation of the results suggests a role of Gal4 in nucleic acid binding. Accordingly, the transcriptional activator protein Purbeta [21] and MAP kinaseactivated protein kinase 3 [22] (MAPKAPK3), which plays an essential role in modulating gene transcription in the nucleus, are among the top upregulated proteins.
Other signaling proteins involved in transcriptional control include basic transcription factor 3 (BTF3) [23] and breast cancer anti-estrogen resistance 1 (BCAR1) protein [24], which are downregulated upon Gal4 expression. For validation of the proteomic results, we focused on these proteins, since there were hints of their potential function in cancer. For this purpose, we chose combining radiometric metabolic labeling with immunoprecipitation, since this method allows exact quantitative measurements. Moreover, applying this approach to both HCT116-Gal4 clones underscores the validity of the data. Thus, this independent approach corroborates the proteomic data.
MAPKAPK3 is part of the p38 mitogen-activated protein kinase (p38MAPK) pathway. It has a high structural identity with MAPKAP2 and shares the same substrate specificity, suggesting very similar functions in biological systems. MAPKAP2 is considered a master regulator of RNA-binding proteins. However, the specific functions of MAPKAPK3 are not clearly defined yet [25]. Tumorigenic and also tumor-suppressive functions of MAPKAPK3 have been described. In cancer cell models, its loss provided a growth advantage [26]. BTF3 has been described as an oncogene in various cancers. In CRC tissue, BTF3 is overexpressed and correlates with a poor prognosis [27]. A recent study indicates that the oncogenic action in colon cancer is linked to an induction of a DNA helicase (CHD1L) and inhibition of E3 ubiquitin ligase HERC2-mediated p53 degradation [28]. BCAR1 belongs to the family of CAS proteins, which are overexpressed in many solid tumors, where they promote invasion and metastasis [29,30].
The most pronounced effect observed upon reconstituted Gal4 expression was an approximately 10-fold upregulation of ZBTB7A phosphorylation at serine residue 549. The protein acts as a pleiotropic transcription factor that is functionally involved in various stages of cell proliferation and differentiation. In addition, ZBTB7A overexpression has also been linked to tumorigenesis and metastasis in various cancer types, including breast, prostate, lung, ovarian and colon cancer [31]. The S549 phosphorylation site has already been described [32], but its physiological significance is not yet known. The new finding of its prominent Gal4-dependent regulation points to the possibility that it might be involved in the control of ZBTB7A activity, which requires future functional studies. FOXK1 is a transcription factor that regulates the expression of many genes, thereby controlling various cellular functions, including cell cycle, cell growth, proliferation, apoptosis, autophagy, stress resistance, metabolism, DNA damage, drug resistance, angiogenesis and carcinogenesis [33]. FOXK1 is a substrate of glycogen synthase kinase 3 (GSK3), and the phosphorylation of FOXK1 regulates its translocation between the cytoplasm and nucleus. FOXK1 carries a variety of phosphosites that may be involved in its regulation (www.phosphonet.ca, accessed on 1 April 2022). The phosphorylation of three sites in FOXK1 (T436, S441 and S445) is upregulated by Gal4. In a study that quantitated all phosphosites of FOXK1 during mitotic cell proliferation, all three sites were already listed, but they were not changed during mitosis [34]. CK2beta is the regulatory subunit of the tetrameric CK2 protein complex that is involved in various cellular processes, with a focus on the control of cell growth, proliferation and survival. Its activity is increased in nearly all cancers, and upregulated CK2 is considered a key player in cancer biology [35]. CK2beta modulates the enzymatic activity and substrate specificity, as well as the assembly of the CK2 complexes [36]. In turn, CK2beta is regulated by phosphorylation/autophosphorylation [37,38]. However, the influence of phosphosite S228 is not yet defined. The expression of transcriptional activator protein PURB was upregulated by Gal4, while its phosphorylation at sites S6 and S8 was decreased. It belongs to the PUR protein family that are involved in the initiation of DNA replication, cell cycle regulation and in the regulation of transcription and RNA translation [39]. The expression or activity of PURB is altered in several cancers [40,41]. Its phosphorylation at S6 and S8 was described in a previous study [42]. The observation that an increased expression of PURbeta correlates with decreased phosphorylation might be considered as a hint that phosphosites S6 and S8 are involved in the Gal4-controlled stability of PURB.
Altogether, our cell line model allows the targeted analysis of intracellular Gal4 functions, as exemplified by our proteomic and phosphoproteomic analyses. Our study suggests a nuclear function of Gal4 in transcription control and defines promising candidates for future research. However, each candidate-in particular, those that have been confirmed by metabolic labeling/immunoprecipitation-has to be pursued in extensive follow-up studies. Nevertheless, the present study highlights the potential of our experimental strategy to set starting points and to direct future research on a specific protein with potentially important but still poorly defined functions, such as Gal4. It may be adapted to any other protein of interest.

Cancer Cell Lines and Culture Conditions
Colorectal cancer cell lines were grown under the standard conditions in DMEM/F12 medium (Thermo Fisher Scientific, Darmstadt, Germany) supplemented with 10% FCS in the presence of 100 U/mL penicillin and 100 µg/mL streptomycin (PAA Laboratories GmbH, Cölbe, Germany). HCT116 (ECACC#91091005), Colo 205 (ECACC#87061208) and HeLa (ECACC#93021013) cells were obtained from the European Collection of Authenticated Cell Cultures (ECACC; https://www.phe-culturecollections.org.uk (accessed on 7 January 1997). HCT116 HygTK cell line was generated previously in our laboratory [15]. All cell lines were regularly checked for mycoplasm contamination using the Mycoplasma free IAA was blocked by adding 5 µL of 1M DTT-containing solution and incubation for 15 min at 37 • C. Pre-digestion was performed with 2.5 µg Lys-C (Promega GmbH, Heidelberg, Germany) in 100 mM NH 4 HCO 3 overnight incubation at 37 • C with constant gentle shaking. Then, the samples were diluted with 700 µL of 100 mM NH 4 HCO 3 and 0.125 mM CaCl 2 to achieve the UREA concentration (~1 M) required for the perseverance of tryptic activity. Tryptic digestion was conducted overnight at 37 • C with 6.5 µg trypsin (Promega GmbH, Heidelberg, Germany). After digestion, 10 µg of the protein digested were subjected to a shotgun mass spectrometry analysis, while 400 µg protein digest underwent phosphopeptide enrichment for a phosphorylation analysis.

Phosphopeptide Enrichment
In order to clean and concentrate the peptide mixtures after tryptic digestion, the StageTip procedure [45] was applied using C18 material and the reversed phase material (Oligo™ R3, Thermo Fisher Scientific, Darmstadt, Germany) packed into a pipette tip (volume up to 200 µL). Briefly, binding was performed with 2.5% formic acid, followed by washing with 2.5% formic acid and elution with 2 times 100 µL of 0.6% acetic acid in 80% acetonitrile. The samples were dried in a vacuum centrifuge.
Each sample was then resuspended in 300 µL binding buffer (80% acetonitrile (ACN), 5% trifluoroacetic acid (TFA) and 0.2% glycolic acid) by sonication and 30 min of incubation on a Thermoshaker (250 rpm, RT) and added to 100 µL of freshly prepared 5% PureCube Fe 3+ -NTA MagBeads (31505-Fe; Biozym Scientific GmbH, Hessisch-Oldenburg, Germany) in the binding buffer. Samples were incubated with the beads for 1.5 h on a rotator at RT. Subsequently, the beads were washed twice with 80% ACN/1% TFA, followed by a wash with 10% ACN/0.2% TFA. Phosphopeptides were eluted by 30 min of incubation with 80 µL of 1% ammonium hydroxide (250 rpm, RT). The samples were then purified by the StageTip procedure, as described above. Each solution was dried completely in a vacuum centrifuge and frozen.

LC-MS/MS
Dried peptides were reconstituted prior to LC-MS/MS analysis in 2.5% Hexafluoro-2-propanol, 0.1% TFA in water or 50 mM citrate and 0.1% TFA, respectively. First, the peptides were loaded onto a trapping cartridge (Acclaim PepMap300 C18, 5µm, 300 Å wide pore, Thermo Fisher Scientific, Darmstadt, Germany) and desalted for 3 min using 0.05% TFA. Peptide separation was performed using a multistep gradient of buffer A (0.1% formic acid in water) and buffer B (0.1% formic acid in acetonitrile), with the main step ramping up the buffer B concentration from 5% to 30% (28% for phosphopeptides) over 132 min on a nanoEase MZ Peptide analytical column (300 Å, 1.7 µm, 75 µm x 200 mm, Waters) using an UltiMate 3000 UHPLC system (60 min of total analysis time). Eluting peptides were subsequently analyzed by an Orbitrap Exploris 480 mass spectrometer (Thermo Fisher Scientific, Darmstadt, Germany) running in data-dependent acquisition mode, where one full scan (380-1400 m/z, 300% AGC target, maxIT 45 ms) at 120-k resolution was followed by MS/MS scans for 2-s cycle times. Precursors were isolated with 1.4 m/z (1.2 m/z), peptides were fragmented using 26 NCE (28 NCE) and MS/MS scans were recorded at 15-k resolution (100% AGC target, maxIT 22 ms; 200% AGC, maxIT 54 ms). Unassigned and signals were excluded from fragmentation, and dynamic exclusion was set to 35 s for 2-6× charged features.

Protein and Phosphopeptide Identification and Quantification
The MS files were processed with MaxQuant software (version 1.6.14; Max Planck Institute of Biochemistry, Munich, Germany) [46] and searched with the Andromeda search engine [47] against the human UniProt database (sequences 74,830 entries). Enzyme specificity was set to that of trypsin, allowing for cleavage of the N-terminal to proline residues and up to 2 missed cleavage sites (for proteome) or up to 4 missed cleavage sites (for phosphopeptides). A minimum peptide length of 7 amino acids was required. Carbamidomethylation (C) was set as a fixed modification, whereas oxidation (M), deamidation (NQ), protein N-terminal acetylation and, if necessary, phosphorylation (STY) were considered as variable modifications. No labeling or double SILAC labeling was defined according to a maximum of 3 or 5 labeled amino acids. Mass tolerances were defined for the precursor and fragmented ions as follows: MS first search-20 ppm, MS main search-6 ppm and MS/MS deisotoping tolerance-0.7 ppm. The false discovery rates (FDRs) at the protein and peptide levels were set to 1%. SILAC-based quantification was based on unique and razor peptides only, and a minimum of two ratio counts was required. Peptide ratios were calculated and normalized for each arginine-and/or lysinecontaining peptide as described [46]. In addition, the "match between the runs" feature was implemented with default settings to increase the number of quantified peptides.
Further data analysis was performed in Perseus (version 1.6.15; Max Planck Institute of Biochemistry, Munich, Germany). Matches to the reverse database proteins identified by one site only in modified peptides and common contaminants (KRT2 and KRT10) were removed from the MaxQuant output. Exclusively, phosphosites quantified in at least 2 (out of 3) replicates and with a localization probability higher than 0.75 were subjected to further analysis. Only proteins identified with at least two unique peptides and quantified in at least 2 (out of 3) biological replicates were considered for the subsequent analysis. The obtained phosphopeptides ratios were corrected for differential protein expression by dividing the matched protein ratios. Proteins and phosphosites changed by >1.5-fold in at least two biological replicates were considered regulated. In addition, to each identified and quantified phosphosite, information from the PhosphoSite Plus database [48] was assigned, including known phosphosites.

Data Analysis
A global interaction network of regulated proteins and phosphoproteins was predicted in STRING v11.5 (available at: www.string-db.org (accessed on 1 April 2022)) [49]. Each protein-protein interaction (PPI) has a combined score (edge score), which represents the reliability of the interaction between proteins. Settings included (i) 40 seed proteins, (ii) human genome interactions, (iii) a highest confidence PPI interaction score of 0.9 and (iv) 500 first shell nodes, resulting in a PPI-enrichment p-value of <1 × 10 −16 . In addition, an enrichment analysis of the regulated proteins and phosphoproteins were also performed in STRING v11.5 for the Gene Ontology Cellular Compartments (GOCC) and Molecular Function (GOMF). Cluster analysis was based on k-means clustering and restricted to 8 clusters. Multiple hypothesis testing was controlled by using a Benjamini-Hochberg FDR. Visible clusters on PPI maps were assigned by color coding the nodes.

Metabolic Labeling and Immunoprecipitation of Proteins
For the metabolic labeling of cellular proteins, cells were grown in the presence of [ 35 S]methionine and [ 32 P]-orthophosphate to test the protein phosphorylation, as described previously [9].
Immunoprecipitation was performed using protein A/G-presenting magnetic beads (Thermo Fisher Scientific, Darmstadt, Germany) as described before [9] using 1 to 2 mg of cell protein lysate in RIPA buffer with appropriate amounts of the following antibodies