Metabolic clogging of mannose triggers genomic instability via dNTP loss in human cancer cells

Mannose has anti-cancer activity that inhibits cell proliferation and enhances the efficacy of chemotherapy. How mannose exerts its anti-cancer activity, however, remains poorly understood. Here, using genetically engineered human cancer cells that permit the precise control of mannose metabolic flux, we demonstrate that the large influx of mannose exceeding its metabolic capacity induced metabolic remodeling, leading to the generation of slow-cycling cells with limited deoxyribonucleoside triphosphates (dNTPs). This metabolic remodeling impaired dormant origin firing required to rescue stalled forks by cisplatin, thus exacerbating replication stress. Importantly, pharmacological inhibition of de novo dNTP biosynthesis was sufficient to retard cell cycle progression, sensitize cells to cisplatin, and inhibit dormant origin firing, suggesting dNTP loss-induced genomic instability as a central mechanism for the anti-cancer activity of mannose.


Introduction:
In mammals, mannose is a monosaccharide that is essential for life and is synthesized de novo from glucose through a glycolysis branch ( fig. S1A). This process requires the action of mannose phosphate isomerase (MPI), the enzyme that catalyzes the interconversion between fructose-6-phosphate (Fruc-6-P) and mannose-6-phosphate (Man-6-P) (Alton et al., 1998). Man-6-P is further converted to GDP-mannose for the biosynthesis of asparagine-linked 5 glycans (N-glycans) in the endoplasmic reticulum (Harada et al., 2013). In the secretory pathway, extensive trimming of N-glycans by mannosidases generates free mannose, which is secreted from cells and contributes to the extracellular pool of mannose (30-130 μM in blood) (Sharma and Freeze, 2011). The extracellular mannose can be taken up by cells and the salvaged mannose is directly converted to Man-6-P, the majority of which is efficiently directed into 10 glycolysis by the action of MPI for unknown reasons (Ichikawa et al., 2014;Sharma and Freeze, 2011).
The large influx of mannose is known to suppress cell proliferation and enhance the efficacy of chemotherapy, particularly in cancer cells that express low levels of MPI (Gonzalez et al., 2018), although the underlying mechanisms remain poorly understood. It has been known for nearly a 15 century that feeding mannose to honeybees, which are believed to express negligible amounts of MPI (Sols et al., 1960), is lethal to these insects (Staudenmayer, 1939). This is because the intracellular levels of Man-6-P exceed the capacity to metabolize it, and the excess Man-6-P inhibits glucose metabolism and decreases the intracellular ATP pool (DeRossi et al., 2006;Sols et al., 1960). This metabolic deficiency is called honeybee syndrome; however, it is unknown 20 whether this syndrome plays a key role in the anti-cancer activity of mannose, and if so, what metabolic checkpoints are targeted by mannose to trigger its anti-cancer activity. Moreover, it is enigmatic how mannose sensitizes poorly proliferating cancer cells to chemotherapy that is designed to target actively proliferating cells. In the present study, we addressed these two

Induction of honeybee syndrome suppresses cell proliferation and increases chemosensitivity
To establish MPI-KO human cancer cells using the CRISPR-Cas9 system, we exploited the mannose auxotrophy and sensitivity observed in MPI-KO mouse embryonic fibroblasts  KO MEFs) (DeRossi et al., 2006). The addition of a physiological concentration of mannose (50 μM, unchallenged) to culture medium supported the proliferation of MPI-KO MEFs ( fig. S1B).
In contrast, mannose starvation or the addition of a supraphysiological concentration of mannose (5 mM, challenged) suppressed the proliferation of MPI-KO MEFs, but not that of wild-type MEFs ( fig. S1C). On these bases, we knocked out the MPI gene in human fibrosarcoma HT1080 10 cells and screened the gene-edited clones under mannose-unchallenged conditions. Three MPI-KO HT1080 clones were obtained (Fig. 1A). In one clone (#1), cell division stopped under mannose-unchallenged conditions, but the other two clones (#2 and #3) could proliferate. These two clones exhibited mannose auxotrophy and sensitivity as expected ( Fig. 1B and fig. S2A), while the parental HT1080 cells showed marginal defects in cell proliferation at mannose 15 concentrations higher than 15 mM (Fig. 1C). The effects of mannose starvation and mannose challenge on the proliferation of MPI-KO cells were almost fully rescued by reintroduction of the human MPI gene (Fig. 1D-F and fig. S2B-D), ruling out the potential off-target effects of gene editing. As expected, mannose challenge to MPI-KO HT1080 cells caused the dramatic accumulation of hexose-6-phosphate (the sum of glucose-6-phosphate and Man-6-P) compared 20 with that under mannose-unchallenged conditions (Fig. 1G). Consistent with the essential role of exogenous mannose in GDP-mannose production for N-glycan biosynthesis in MPI-KO MEFs (Harada et al., 2013), mannose starvation severely decreased N-glycosylation in MPI-KO HT1080 cells (Fig. 1H). However, mannose challenge showed negligible effects on N-glycosylation, indicating that mannose challenge suppresses cell proliferation through a mechanism distinct from N-glycosylation defects.
Mannose challenge increased the sensitivity of MPI-KO HT1080 cells to DNA replication inhibitors (i.e., cisplatin and doxorubicin) when the cells had been preconditioned with excess mannose prior to the drug treatment (Fig. 1I, J and fig. S2E, F). We also generated  HeLa cells as another cell model ( fig. S3A-I), and found that mannose challenge also increased the sensitivity of these cells to cisplatin and doxorubicin ( fig. S3J, K). All of these results demonstrate that the induction of honeybee syndrome suppresses cell proliferation and increases chemosensitivity in our MPI-KO human cancer cell models. We mostly used MPI-KO HT1080 (#3) cells for subsequent study, while the other cell models showed similar results.

Mannose challenge generates slow-cycling cells
Cell proliferation is tightly controlled by cell cycle progression. To explore the mechanism behind the anti-proliferative activity of mannose, we compared the cell cycle progression between mannose-challenged and unchallenged MPI-KO HT1080 cells by using a two-color 15 fluorescent ubiquitination-based cell cycle indicator [Fucci(CA)] (Sakaue-Sawano et al., 2017). In this reporter system, G1 phase was defined by the exclusive expression of mCherry-hCdt1(1/100)Cy(-), which was rapidly turned off upon the onset of S phase where mVenus-hGem(1/110) gradually accumulated ( Fig. 2A, B and movie S1). The mCherry-hCdt1(1/100)Cy(-) was re-expressed upon the onset of G2 phase (the double-positive phase), 20 followed by the termination of M phase where mVenus-hGem(1/110) was abruptly turned off.
Under mannose-unchallenged conditions, MPI-KO HT1080 cells showed exponential growth ( Fig. 2C) and a typical Fucci(CA) signal profile (Fig. 2B), as reported in HeLa cells (Sakaue-proliferation ( Fig. 2C) and significantly prolonged the cell cycle, showing a variety of atypical Fucci(CA) signal profiles (Fig. 2E, F, fig. S4). We classified these profiles based on the order of expression of Fucci(CA) reporters [mCherry-hCdt1(1/100)Cy(-) as R, mVenus-hGem(1/110) as G, and the double-negative phase as Dn, fig. S4]. The classification analysis revealed that a small 5 proportion of the mannose-challenged cells showed normal-like but strikingly extended Fucci(CA) signal profiles (Fig. 2E, F and fig. S4, normal-like, 12% of the total). Notably, these normal-like cell populations frequently failed to undergo cytokinesis in M phase and re-entered G1 phase without generating two equivalent daughter cells (movie S2). Other fractions included cells that were arrested in G1 phase (Fig. 2E, F and fig. S4, RRR,26.7%), that showed little to 10 no double-positive phase (i.e., the G2-to-M phase) ( Fig. 2F and fig. S4, RGR,15.8% and GRG,9.9%, GDn, 5.9%, RGG, 5.0%, RGDn, 4.0%; 40.6% of total), or that remained in the doublenegative phase (Fig. 2E, F and fig. S4, Dn,8.9%), which is normally seen only at the G1/S transition (Sakaue-Sawano et al., 2017). These results suggest that mannose challenge severely impairs the entry of the cells into S phase and its progression to mitotic phase. Strikingly, 15 however, switching of the mannose-challenge medium to the mannose-unchallenged medium after long-term mannose challenge (6 days) resulted in robust cell proliferation (Fig. 2G), suggesting that some fraction of the mannose-challenged cells had entered a quiescent state upon mannose challenge. Supporting this assumption, mannose challenge induced early accumulation of the cell cycle inhibitors p21 and p27 (Abukhdeir and Park, 2008) (Fig. 2H) and decreased cell 20 populations that were actively incorporating 5-bromo-2'-deoxyuridine (BrdU) into DNA (Fig.   2I, J). Small proportions of the mannose-challenged cells exhibited DNA content greater than 4n after 2-day mannose challenge, suggesting that these cell populations underwent endoreplication (Edgar and Orr-Weaver, 2001). Collectively, these results indicate that mannose challenge suppresses cell proliferation through a complex mechanism involving extremely slow cell cycle progression.

Mannose challenge limits DNA synthesis at ongoing replication forks
We adopted a proteomic approach to dissect the molecular mechanism by which mannose 5 challenge generated slow-cycling cells in MPI-KO HT1080. Of over 7,000 proteins identified in the proteomic datasets (data S1), proteins that were significantly up-and downregulated relative to those in the cells cultured for 1 day under mannose-unchallenged conditions were extracted at each time point (data S2). Functional annotation analysis of the proteomic data using DAVID bioinformatic resources (https://david.ncifcrf.gov) revealed the downregulation of proteins 10 related to the cell cycle and DNA replication in mannose-challenged cells (Fig. 3A), which were enriched with the mini-chromosome maintenance 2-7 (MCM2-7) complex (Fig. 3B). Western blot analysis and quantitative polymerase chain reaction (qPCR) confirmed the decrease in the expression levels of MCM2-7 proteins (Fig. 3C) and their genes (Fig. 3D) during mannose challenge over 6 days. The MCM2-7 complex is a core component of DNA helicase that 15 unwinds the DNA duplex at replication forks in S phase (Jones et al., 2021;Rzechorzek et al., 2020), and the complex also plays a central role in the licensing of replication origins in G1 phase by forming a pre-replicative complex with the six-subunit origin recognition complex (ORC1-6), CDC6, and CDT1 on chromatin (Frigola et al., 2017;Remus et al., 2009;Zhai et al., 2017). CDC6 and CDT1, which were not detected in our proteomic analysis, were depleted after 2-day 20 mannose challenge, whereas the ORC2 subunit was expressed at relatively constant levels (Fig.   3C). These results indicate that mannose challenge induces slow proteomic alterations in origin licensing factors, while this cannot be an essential trigger for the generation of slow-cycling cells, as these cells appeared immediately after mannose challenge.
We performed chromatin flow cytometry for MCM2 to elucidate the impact of mannose challenge on the chromatin-bound states of the MCM complex. Under the mannoseunchallenged conditions, the binding of MCM2 to chromatin peaked at the G1/S boundary in 5 MPI-KO HT1080 cells and MPI-KO HeLa cells (Fig. 3E), indicating that replication origins were fully licensed (Matson et al., 2017). The cells with the fully licensed origins entered S phase as a tight population, and MCM2 dissociated from chromatin as two replication forks converged and terminated ( Fig. 3E) (Low et al., 2020). Mannose challenge did not severely impair origin licensing (Fig. 3E), but the same treatment caused the accumulation of MCM2-10 positive chromatin in S phase (Fig. 3E). Notably, the mannose-challenged cells were not actively incorporating BrdU into DNA (Fig. 2I, J and fig. S5A), suggesting that ongoing replication forks were stuck on chromatin with little DNA synthesis under the mannose-challenged conditions. Taking these findings together, the limited DNA synthesis at ongoing replication forks likely contributes to the abnormally extended progression of S phase in the mannose-15 challenged cells.

Mannose challenge disengages dormant origins from DNA synthesis during replication stress
Although our findings indicated that mannose challenge limits DNA synthesis at ongoing 20 replication forks, its relevance to the increased chemosensitivity was still unclear. In humans, replication origins are licensed far more than actually used for DNA replication (Langley et al., 2016). The excess origins remain dormant under physiological conditions, while the dormant origins are activated for DNA replication when nearby replication forks stall upon encountering DNA lesions, thus preventing cells from the permanent replication arrest that leads to cell death (Blow et al., 2011;Ge et al., 2007;Kawabata et al., 2011;Shima et al., 2007). To test whether mannose challenge may also impair DNA synthesis from dormant origins during replication stress, we compared BrdU incorporation between the cells that were first pulsed with cisplatin to 5 induce replication stress, followed by being left untreated or being treated with an inhibitor of the ataxia telangiectasia and Rad3-related protein (ATR) to forcibly activate dormant origins (Moiseeva and Bakkenist, 2019; Moiseeva et al., 2019) (Fig. 4A). In the mannose-unchallenged MPI-KO HT1080 cells (Fig. 4B, C) and MPI-KO HeLa cells ( fig. S5A, B), the cisplatin treatment alone partially suppressed the incorporation of BrdU, which was greatly recovered by 10 ATR inhibition with VE-821 (ATRi) (Charrier et al., 2011;Prevo et al., 2012;Reaper et al., 2011), indicating that dormant origins are present in excess and their activation can engage in DNA synthesis during replication stress. In the mannose-challenged cells, however, the cisplatin treatment more severely reduced BrdU incorporation, which was barely restored by ATRi treatment (Fig. 4B, C and fig. S5A, B). Forced activation of dormant origins during replication 15 stress is known to cause the uncoupling of DNA unwinding and synthesis (Murai et al., 2018), leading to the accumulation of single-stranded DNA on chromatin marked by phosphorylation of the replication protein A2 (RPA2) (Fig. 4D), confirming the activation of dormant origins by ATRi treatment in both mannose-challenged and -unchallenged cells. These results indicate that mannose challenge limits DNA synthesis from dormant origins during replication stress.

20
The replication initiation step requires chromatin loading of cell division cycle 45 (CDC45), one of the essential and limiting factors for replication initiation (Moyer et al., 2006), while this process was not severely impaired by mannose challenge (Fig. 4D and fig. S6A, B). Under mannose-unchallenged conditions, cisplatin treatment induced the unloading of CDC45 from chromatin, which was restored by the forced activation of dormant origins ( Fig. 4D and fig.   S6A, B). However, mannose challenge abrogated the CDC45 unloading/reloading dynamics under replication stress conditions ( Fig. 4D and fig. S6A, B). The deficiency in the DNA synthesis from dormant origins and the CDC45 dynamics was associated with the increase in cisplatin-induced gH2AX (Fig. 4D, E and fig. S5C), the phosphorylated form of the histone 5 H2AX that marks DNA double-strand breaks (Kuo and Yang, 2008). The endogenous levels of gH2AX were detected in S phase of both mannose-challenged and -unchallenged cells (Fig. 4F,   fig. S5D and S6C), and cisplatin strongly induced gH2AX throughout S phase even in the mannose-challenged cells ( Fig. 4F and fig. S5D). Together, these results indicate that mannose challenge disengages dormant origins from DNA synthesis during replication stress, thus 10 exacerbating DNA damage.

Mannose challenge generates a distinct metabolic landscape
Although our data showed that mannose challenge generated slow-cycling cells that failed to engage dormant origins in DNA synthesis during replication stress, the underlying mechanism 15 that functionally links these two phenotypes was still unclear. Since cell cycle progression and DNA replication are metabolically demanding processes (Heinemann and Zylstra, 2022), we hypothesized that the metabolic checkpoints activated by honeybee syndrome (DeRossi et al., 2006;Sols et al., 1960) may play a key role in the anti-cancer activity of mannose. To test this hypothesis, we first examined the impact of mannose challenge on the cellular bioenergetics in 20 MPI-KO HT1080 cells by monitoring the real-time changes in glycolytic flux in the form of extracellular acidification rate (ECAR) and the activity of oxidative phosphorylation (OXPHOS) in the form of oxygen consumption rate (OCR). Mannose challenge increased OCR in response to a steep drop of ECAR (Fig. 5A, B). This bioenergetic compensation was not sufficient to maintain the ATP production rates in full (Fig. 5C). Consistent with this, mannose challenge decreased the ATP pool in MPI-KO HT1080 cells, MPI-KO HeLa cells, and MPI-KO MEFs ( Fig. 5D, E and fig. S7), and the remaining pool was completely depleted by co-treatment with IACS-010759, a preclinical small-molecule inhibitor of complex I of the mitochondrial 5 respiratory chain (Molina et al., 2018) (Fig. 5D, E and fig. S7), which in turn increased necrosis (Fig. 5F). These results indicate that, in honeybee syndrome, the bioenergetic balance is impaired by glycolysis being acutely shut down and OXPHOS being boosted.
To provide deeper insights into the metabolic landscape generated by honeybee syndrome, we compared the metabolome between mannose-challenged and -unchallenged MPI-KO HT1080 10 cells ( fig. S8A and data S3). One-day mannose challenge greatly increased Fruc-6-P, while the three-carbon metabolites in the lower glycolysis chain, including lactate, were substantially decreased ( Fig. 5G), supporting our earlier findings that mannose challenge decreased ECAR to the basal level. Despite this glycolytic insufficiency, the amounts of metabolites in oxidative and non-oxidative arms of the pentose phosphate pathway (PPP) remained relatively unchanged in 15 the mannose-challenged cells, except for the large accumulation of 6-phosphogluconoate and the slight decrease in sedoheptulose-7-phosphate ( Fig. 5H). In contrast, mannose challenge severely decreased tricarboxylic acid (TCA) cycle intermediates (isocitrate, 2-oxoglutarate, succinate, fumarate, and malate) that are used to generate NADH and FADH2 as electron donors for the mitochondrial respiratory chain ( fig. S8B), suggesting their large consumption by the boosting of 20 OXPHOS in honeybee syndrome. Strikingly, the amounts of nucleoside diphosphates and triphosphates, but not nucleoside monophosphates, were proportionally and moderately decreased ( Fig. 5I-K), whereas deoxyribonucleoside triphosphates (dNTPs) were substantially decreased in the mannose-challenged cells (Fig. 5L). Collectively, these results indicate that mannose challenge generates a distinct metabolic landscape in MPI-KO HT1080 cells, which ultimately lead to the depletion of dNTP pools.
Pharmacological inhibition of de novo dNTP biosynthesis retards cell cycle progression, increases chemosensitivity, and inhibits DNA synthesis from dormant origins 5 dNTPs are the essential donor substrates for DNA synthesis, raising the possibility that the dNTP loss caused by mannose challenge may be a major mechanism linking the generation of slowcycling cells and the failure to engage dormant origins in DNA synthesis during replication stress. To test this hypothesis, we inhibited de novo synthesis of dNTPs with hydroxyurea (HU), a highly potent inhibitor of ribonucleotide reductase (RNR) (Elford, 1968), and found that the 10 combination of HU and cisplatin more severely reduced cell viability than cisplatin treatment alone, independently of the presence or absence of the MPI gene (Fig. 6A). This chemosensitizing effect of HU was associated with the strong induction of gH2AX (Fig. 6B).
HU is widely used as an agent to arrest cells in S phase by depleting dNTP (Davis et al., 2001).
Profiling of the chromatin-bound MCM2 showed that HU treatment at a moderate concentration 15 (0.25 mM) resulted in the arrest of cells in the early to late S phase, while a higher concentration of HU (1 mM) almost completely arrested the cells at the G1/S boundary (Fig. 6C), partly recapitulating the inhibitory effects of mannose challenge on cell cycle progression. As with the case of mannose challenge, the forced activation of dormant origins by ATRi failed to increase BrdU uptake in the presence of HU ( Fig. 6D-G), indicating that the DNA synthesis from 20 dormant origins is highly sensitive to the pool size of dNTPs. Unlike mannose challenge, however, HU treatment did not severely impair the CDC45 unloading/reloading dynamics during disengagement of dormant origins from DNA synthesis.

Discussion
MPI is the sole enzyme that catalyzes the interconversion between Fruc-6-P and Man-6-P in mammals. The conversion of Fruc-6-P to Man-6-P mediated by MPI is central to the synthesis of GDP-mannose from abundant glucose for normal glycosylation. In contrast, the conversion of Man-6-P to Fruc-6-P mediated by the same enzyme is critical to directing the excess Man-6-P to 5 glycolysis, while the cellular function of this seemingly wasteful metabolic pathway has long remained unknown. In the present study, we employed MPI-KO human cancer cells to explore the key mechanism behind the anti-cancer activity of mannose, and demonstrated that the large influx of mannose exceeding the capacity to metabolize it, that is, the onset of honeybee syndrome, induced dramatic metabolic remodeling that led to dNTP loss. These cells were 10 cycling extremely slowly, and upon encountering replication stress, they were unable to rescue stalled forks via dormant origins, thus exacerbating replication stress. These findings shed light on the conversion of Man-6-P to Fruc-6-P mediated by MPI as a genome guard through the maintenance of dNTP homeostasis, which could be threatened by dietary mannose intake.
Our findings demonstrate that the metabolic landscape formed by honeybee syndrome is far 15 more complex than previously proposed (DeRossi et al., 2006;Sols et al., 1960). Acute shutdown of glycolysis is the hallmark of honeybee syndrome, which concomitantly boosts OXPHOS and forces the cellular metabolome to adapt to this bioenergetic imbalance. Strikingly, the metabolic checkpoints targeted in honeybee syndrome are quite distinct from those expected from in vitro enzymatic activity assays (DeRossi et al., 2006). In particular, despite the fact that 20 Man-6-P is a highly potent inhibitor of glucose-6-phosphate dehydrogenase in vitro (DeRossi et al., 2006), PPP metabolites were shown not to be depleted in honeybee syndrome, suggesting that previously unidentified mechanisms are operating to cause dNTP loss. In addition to the essential role of dNTPs in the normal progression of S phase, accumulating evidence that the In conclusion, we demonstrate that honeybee syndrome plays a critical role in the anti-cancer activity of mannose, at least in part through dNTP loss. Ensuring appropriate amounts of dNTP is essential for normal DNA replication, as well as for the recovery from replication stress, thus 20 providing a plausible mechanism for how mannose sensitizes poorly proliferating cancer cells to chemotherapy.

Experimental models
The human fibrosarcoma cell line HT1080 used in this study was our laboratory stock and validated by short tandem repeat profiling (Promega). The human cervix adenocarcinoma cell The plasmids used in this study are listed in Supplementary Table S2. The coding sequence of the human MPI gene was amplified by PCR using Phusion High-Fidelity DNA polymerase (M0530, New England BioLabs), a primer set (Supplementary Table S1), and cDNA from HuH-7 cells as a template. The PCR products were purified from agarose gels using NucleoSpin Gel and PCR Clean-Up Kit (740609, Clontech) and subcloned into the pENTR/D-TOPO vector 5 (K240020, Thermo Fisher Scientific), in accordance with the manufacturer's instructions. The yielded plasmid (pENTR-hMPI) was used as a template to amplify the MPI gene by PCR using a primer set (Supplementary Table S1) and the PCR products were cloned into pMXs-Neo retroviral expression vector (RTV-011, Cell Biolabs) to yield pMXs-Neo-hMPI by using In-Fusion HD Cloning Kit (639648, Clontech), in accordance with the manufacturer's instructions.  Table S1) and cloned into pMXs-Neo to yield pMXs-Neo-mCherry-hCdt1(1/100)Cy(-) using the In-Fusion HD Cloning Kit. pMMLV-mVenus-hGem(1/110):IRES:Bsd was constructed and purchased from Vector Builder.

Chromatin extraction
Cells were washed once with PBS, scraped in the same buffer, and pelleted by centrifugation at 200 ´ g for 5 min at 4°C. The wet cell weight was measured after the supernatant had been completely removed. The cell pellet was extracted by incubating for 5 min at 0°C at a concentration of 10 mg wet cells/100 μL of CSK buffer [10 mM Hepes-KOH, pH7.4, 340 mM sucrose, 10% (v/v) glycerol, 10 mM KCl, 1.5 mM MgCl2, 0.1% (v/v) Triton X-100, 1 mM ATP, 1 mM DTT, 1 ´ cOmplete Protease Inhibitor Cocktail, and 1 ´ PhosSTOP]. The homogenate was centrifuged at 1,390 ´ g for 5 min at 4°C and the pellet was resuspended in 100 µL of CSK 5 buffer. The homogenate was centrifuged again at 1,390 ´ g for 5 min at 4°C. The pellet was resuspended in 100 μL of CSK buffer and mixed with 100 µL of 2 ´ Laemmli sample buffer containing 10% b-mercaptoethanol. The samples were homogenized by two cycles of 10-s sonication at 0°C using a probe-type sonicator (VP-5S, TAITEC) and the homogenates were denatured by heating at 100°C for 3 min.   (Fig. 2E, fig. S4) were visually inspected for classification.
Statistical analysis 15 R version 3.3.3 and Prism 9 were used for statistical analysis. Statistical analysis was performed by applying an unpaired two-sided Welch's t-test for comparison of the means between two groups. Comparisons of the means among more than two groups were performed with one-way or two-way analysis of variance (ANOVA) followed by post hoc testing with Dunnett's test or Tukey's test. Data are reported as the mean ± standard deviation (SD). P-values < 0.05 were 20 considered to be statistically significant.

30
Authors declare that they have no competing interests.

Data and materials availability:
The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD036449 and 10.6019/PXD036449 (Username: reviewer_pxd036449@ebi.ac.uk; Password: aRfYzEvX).

Other Supplementary Materials for this manuscript include the following:
movies S1 to S2 Data S1 to S3 [S1.      Table S1.
List of primers used in this study.