Yeast TIA1 coordinates with Npl3 to promote ATG1 translation during starvation

An in vitro interactome capture reveals several new binding partners of ATG1 mRNA

To identify potential RBPs involved in autophagy regulation at the level of the ATG1 transcript, we performed an in vitro transcribed ATG1 5′ UTR RNA affinity isolation followed by western blot and proteomics to identify interactors in nitrogen-rich (+N) and nitrogen-limited (−N) conditions (Figure 1A). We verified through western blot that Dhh1 and Ded1, known interactors with the ATG1 5′ UTR, were enriched only in the presence of labeled ATG1 5′ UTR RNA (Figure S1A). Proteomics profiling of the affinity isolate identified previously known interacting proteins of the ATG1 5′ UTR, such as Eap1 and Psp2, under nitrogen-starvation conditions (Figure S1B), further validating the utility of this method to identify potential RBPs that could regulate autophagy at the level of ATG1 mRNA. We identified 13 RBPs that were significantly enriched on the ATG1 5′ UTR during nitrogen starvation relative to rich medium (Figure 1B). Out of the 13, we decided to test three non-essential proteins with a previously unknown role in regulating autophagy (Table S1 and Data S1). Accordingly, we created genomic deletion strains of SRO9, NPL3, and MLF3 and tested them for effects on Atg1 protein expression using immunoblotting (Figure 1C). We found that NPL3 deletion severely abrogated the upregulation in Atg1 protein levels relative to the wild-type (WT) strain only under nitrogen-starvation (−N) conditions. Interestingly, npl3 deletion did not affect Atg1 protein levels under growing conditions (+N), suggesting a potential role of Npl3 in regulating Atg1 expression in response to nitrogen starvation. In contrast, the other two deletion strains caused no obvious change in Atg1 levels relative to the WT.

Npl3 is a novel regulator of Atg1 protein expression and autophagy

To confirm that Npl3 does indeed bind to the ATG1 transcript, we performed an Npl3 immunoprecipitation followed by RT-qPCR to measure ATG1 transcript levels and determine whether the transcript was enriched specifically in nitrogen-starvation medium relative to nitrogen-rich medium. Our results show that Npl3 bound to both the 5′ UTR (−290) and the coding region (+35) but not the 3′ UTR (Figure 1D), revealing that Npl3 is indeed an interactor with the ATG1 5′ UTR during nitrogen starvation. We hypothesized that this interaction is crucial for the upregulation of Atg1 protein levels during nitrogen-starvation-induced autophagy. To rule out the possibility that increased protein turnover results in lowered Atg1 protein levels in Npl3 mutants, we performed a cycloheximide (CHX) chase assay in which Atg1 protein levels were allowed to accumulate during 6 h of nitrogen starvation, following which CHX was added. The Atg1 protein levels were then chased in nitrogen-starvation medium for another 1 h and 2 h and measured by immunoblotting using an Atg1-specific antibody. The results show that the turnover of Atg1 protein levels in both WT and npl3Δ strains was similar and, therefore, the regulation was upstream of protein degradation (Figures 1E and 1F). Next, we measured the mRNA levels of the ATG1 transcript in WT and npl3Δ cells to assess whether the lowered amount of Atg1 protein is due to a decrease in ATG1 mRNA levels and, therefore, a form of transcriptional regulation by Npl3. However, both WT and npl3Δ strains had similar levels of ATG1 mRNA transcript in both +N and −N conditions (Figure 1G), suggesting that Npl3 regulates Atg1 protein levels at the post-transcriptional/translational level, only in response to nitrogen starvation.

The decline in Atg1 protein levels upon ablation of the NPL3 gene led us to test whether the decrease affects autophagy activity. Toward this end, we tested GFP-Atg8 processing. Atg8, or GFP-tagged Atg8, is attached to both sides of the phagophore via conjugation to phosphatidylethanolamine. After autophagosome maturation, Atg8/GFP-Atg8 is removed from the autophagosome outer membrane and recycled; in contrast, the protein on the inner surface is delivered to the vacuole after fusion with the autophagosome.¹⁴ Atg8 is degraded within the vacuole lumen, whereas GFP is relatively stable and accumulates within the vacuole. Thus, the generation of free GFP is an indication of autophagy activity. We determined that deletion of NPL3 resulted in decreased GFP-Atg8 processing after 6 h of nitrogen starvation and, therefore, a lower level of autophagy activity (Figures 1H and 1I). To assess whether the lowered autophagic activity is primarily due to a decrease in Atg1 protein levels, we placed the ATG1 open reading frame (ORF) under an inducible promoter to overexpress the Atg1 protein levels and tested whether this overexpression could rescue the autophagy defect measured by GFP-Atg8 processing. Overexpression of Atg1 partially rescued the autophagy defect seen in npl3Δ cells, suggesting that, while ATG1 is a key target of Npl3, there are likely other targets that lead to the autophagy defect (Figures S1C and S1D).

Finally, autophagy is required for yeast cell viability during starvation; autophagy mutants display decreased viability under starvation conditions.¹⁵ Thus, resistance to starvation is another measure of functional autophagy. The npl3 deletion resulted in decreased cellular viability during long-term starvation compared to the WT strain (Figure 1J), although we did note a slight growth defect in growing conditions. This result suggests that Npl3 is an essential factor required for survival during nitrogen starvation and that it primarily acts by promoting Atg1 protein levels and autophagy activity.

Npl3 regulates Atg1 expression and autophagy in an RRM-dependent manner

Npl3 belongs to a family of serine-arginine/SR proteins well studied for its role in mRNA splicing. Npl3 has also been implicated in various stages of gene expression such as transcription elongation, termination, and mRNA splicing. The ATG1 transcript in Saccharomyces cerevisiae does not contain any introns and because the depletion of NPL3 did not affect ATG1 mRNA levels but only the Atg1 protein level, we asked whether (1) export and (2) translational efficiency of ATG1 mRNA was affected. To test whether there was a defect in mRNA export, we isolated total mRNA from nuclear and cytoplasmic fractions from WT and npl3Δ cells after nitrogen starvation and measured the percentage of ATG1 mRNA. The percentage of ATG1 mRNA in the nuclear fraction was significantly higher in npl3Δ cells compared to the WT, whereas there was no significant difference seen with PGK1, suggesting a specific defect in ATG1 mRNA export (Figure 2A).

For this defect to subsequently affect Atg1 protein expression, we hypothesized that the translational efficiency of the transcript should be affected in npl3Δ cells. Accordingly, we assessed the translational efficiency by measuring ribosome occupancy of ATG1 mRNA during nitrogen starvation, carrying out polysome profiling of WT and npl3Δ cells followed by RT-qPCR to determine the localization of the transcript in free and ribosome-bound fractions. The resulting polysome profiles showed a negligible increase in the 60S and 80S fractions; however, there was no significant change in the polysome fraction, suggesting that the global translation remained unaffected in npl3Δ cells during nitrogen starvation (Figure S2A). Higher translational efficiency is a factor of higher ribosome occupancy, which localizes the transcript in the heavier polysome fractions. ATG1 has high ribosome occupancy in WT cells during nitrogen starvation, as evidenced by the recruitment of two or more ribosomes on the transcript (Figure 2B). Furthermore, almost 40% of ATG1 mRNA was occupied by one ribosome, suggesting the importance of monosomes in translating stress-responsive genes.^16,17 In contrast, npl3Δ cells had a significant decrease in the abundance of ATG1 mRNA in the polysome fraction compared to the WT, where two or more ribosomes occupy the transcript. There was a modest yet significant decrease in monosome loading and a proportional increase in the mRNA in the free mRNA-protein complexes (mRNPs) fraction in the npl3Δ strain, suggesting that depletion of NPL3 reduced the ribosome loading onto the ATG1 transcript; this effect was exacerbated in the polysome fractions. In contrast, the localization of the control mRNA, PGK1, was not drastically different in the mutant cells in the polysome fraction, with a modest decrease in the monosome and the free mRNP fractions (Figure S2B). However, the total protein levels of Pgk1 were not affected in npl3Δ cells, suggesting that this difference in PGK1 localization in the various ribosome fractions had no effect on the total protein translation. Together, our data suggest that a decrease in Atg1 protein levels upon npl3 deletion is due a deficiency in ATG1 mRNA export and, therefore, a subsequent decrease in ribosome accessibility of the ATG1 transcript.

Npl3 is an RBP with three potential RNA-binding motifs: two RNA recognition motifs (RRMs) and a C-terminal arginine-serine/RS domain containing an Arg-Gly-Gly (RGG) domain.¹⁸ Genetic interactions analyzing the RNA recognition mechanism of Npl3 highlight the role of RRM1 during chromatin remodeling, while RRM2 might be linked to the regulation of a specific transcript.¹⁹ Furthermore, RRM2 recognizes a 5′-GNGG-3′ motif, which is present in the 5′ UTR of ATG1 starting at position 65 (Figure S2C). To test whether this motif had played a role in the interaction between Npl3 and the ATG1 transcript, we constructed a strain wherein the 5′ UTR of the ATG1 ORF was switched with the 5′ UTR of ATG7, the transcript of which is not a target of Npl3. Using this strain, we performed RNA immunoprecipitation with GFP-tagged Npl3 in WT and Atg1 mutant strains and measured the interaction of the +35 fragment of ATG1. Removal of the 5′ UTR significantly reduced the interaction with the coding region of the ATG1 transcript, suggesting that Npl3 directly targets the 5′ UTR of ATG1 (Figure 2C). Furthermore, we sought to determine whether Npl3 interacted with ATG1 via its RRM2 domain. We deleted the Npl3 residues between 200 and 275 and first tested whether this affected the stability of the protein itself. We found that the truncation had no effect on the total protein stability compared to WT in both +N and −N conditions (Figure S2D); however, we found a modest but significant increase in total protein levels after 6 h of nitrogen starvation compared to growing conditions (Figures S2E and S2F). Next, we tested the effect of this truncation on ATG1 interaction and Atg1 protein levels. We performed RNA immunoprecipitation with protein A (PA)-tagged WT Npl3 and Npl3 ΔRRM2 and analyzed the interaction with ATG1 transcript fragments by RT-qPCR. Deleting RRM2 led to a significant reduction in ATG1 mRNA enrichment compared to WT Npl3 (Figure 2D). Next, we tested the effect of deleting RRM2 on Atg1 protein levels using immunoblotting. Compared to the WT, the ΔRRM2 mutant showed significantly reduced Atg1 protein levels (Figures 2E and 2F). In contrast, ATG1 mRNA levels remained unchanged (Figure 2G), suggesting that the RRM2 motif is essential for Npl3 to promote Atg1 protein expression but not for ATG1 transcription. Finally, we asked whether this reduced binding in the ΔRRM2 mutant led to reduced accessibility for ribosome recruitment on the ATG1 transcript. Indeed, the recruitment of both monosomes and polysomes on the ATG1 transcript was significantly reduced in ΔRRM2 mutants compared to the WT, with a proportional increase in the percentage of ATG1 in the free mRNP fraction (Figure 2G).

When the global translational status was profiled, we noticed a modest decrease in polysomes in the ΔRRM2 mutant during nitrogen starvation (Figure S2F), suggesting potential functional redundancy that may compensate for the loss of the complete protein (i.e., npl3Δ) but not for the RRM2 deletion alone, and further suggesting a weak dominant negative effect of the ΔRRM2 protein, wherein the truncated version of the protein interferes with normal cellular processes, possibly by disrupting interactions or functions typically associated with the full-length protein. Additionally, compared to WT, the control mRNA PGK1 had a similar effect on the monosome and the free mRNP fractions but had no effect on polysomes (Figure S2G). Minor differences in PGK1 localization did not manifest as differences in total Pgk1 protein levels, suggesting that polysome-binding-state differences for PGK1 mRNA in the mutant had no effect on the final translation of the protein. These results suggest that Npl3 interacts with ATG1 mRNA in an RRM2-dependent manner, and this interaction is necessary for the export and the subsequent recruitment of ribosomes and, thus, translational upregulation of Atg1 during nitrogen starvation. Taking the data together, we have identified Npl3 as a novel post-transcriptional regulator of ATG1 mRNA during nitrogen starvation.

Pub1 promotes the translation of ATG1 mRNA and autophagy

Having mapped the interaction network of the ATG1 5′ UTR, we were interested in analyzing the 3′ UTR. The ATG1 3′ UTR region binds to the Pat1-Lsm complex that protects the transcript from decay.²⁰ Specific RBP-3′ UTR interactions also influence association with translational machinery and ribosome components.²¹ We employed the mRNA affinity-isolation approach followed by a proteomics analysis utilizing the labeled 3′ UTR of the ATG1 transcript. Proteomics identification revealed that the ATG1 3′ UTR interacted with the 5′-3′ exonuclease Xrn1, previously shown to regulate Atg1 and autophagy²² (Figure 3A). In addition, we found that the Npl3-interacting protein Gbp2, involved with the nuclear export of transcripts, also associated with ATG1. Hrb1, a paralog of Gbp2, was similarly identified as an interactor of the ATG1 3′ UTR (Data S2).²³ Furthermore, we identified ribosome component Rps27a, suggesting the crucial role of the 3′ UTR in regulating ATG1 mRNA stability, export, and translation (Data S2).

The stress granule core protein Pub1 was also identified as interacting with the 3′ UTR of ATG1 during nitrogen starvation (Figure 3A). Pub1 has not previously been implicated in autophagy, and therefore we were curious as to the significance of the Pub1 interaction with ATG1 mRNA. To confirm this interaction, we immunoprecipitated Pub1-superfolder GFP (sfGFP) expressed from a plasmid, and tested whether it interacted with ATG1 mRNA. We found significant enrichment of the ATG1 3′ UTR with additional binding in the 5′ UTR, suggesting that Pub1 indeed interacted with ATG1 during nitrogen starvation (Figure 3B).

As an interactor of the ATG1 3′ UTR, we next asked whether Pub1 regulates the expression of Atg1. From immunoblot analysis, we determined that pub1Δ cells did not upregulate Atg1 protein levels compared to the WT during nitrogen starvation (Figure 3C). However, the transcript levels of ATG1 remained similar to those of WT under both growing conditions and during nitrogen starvation (Figure 3D). This result suggests that Pub1, similar to Npl3, regulates ATG1 at the post-transcriptional level in response to nitrogen starvation. To confirm this hypothesis, we performed polysome profiling and analyzed the distribution of ATG1 transcript in the different polysome fractions during nitrogen starvation, as we did for the npl3Δ strain. The polysome profiles confirmed that pub1 deletion did not significantly affect the global translation during starvation (Figure S3A). However, pub1 deletion did significantly reduce the localization of ATG1 transcript in the polysome fractions, with a significant increase in the free mRNP fraction, suggesting a failure to recruit polysomes in the absence of Pub1, leading to reduced Atg1 protein levels during nitrogen starvation (Figure 3E).

In contrast, the control RNA PGK1 did not show a drastic change in localization compared to the WT in the polysome fraction, in line with the fact that Pgk1 protein levels were not affected by pub1 deletion (Figure S3B). The difference in ATG1 translation and, therefore, Atg1 protein expression resulted in a difference in autophagy activity as measured by Pho8Δ60 activity (Figure 3F), an enzymatic assay that monitors autophagic flux.²⁴ Similarly, pub1 deletion resulted in a significant difference in GFP-Atg8 processing and free GFP generation, albeit there was a significant difference in total GFP-Atg8 production between WT and pub1Δ (Figures S3C and S3D). When we overexpressed ATG1 under an inducible promoter, the autophagy defect was almost completely rescued, suggesting that the autophagy defect in pub1Δ was primarily due to a deficiency in Atg1 protein levels (Figures 3F, S3C, and S3D). Finally, pub1 deletion drastically reduced the viability of the cells during long-term starvation, providing insights into the critical role of Pub1 in ensuring cell survival during nutrition limitation (Figure 3G).

TIA1 promotes ULK1 expression and autophagy

To explore the conservation of the role of Pub1, a protein we have shown here to regulate Atg1 expression, we studied the capacity of TIA1, its mammalian counterpart, to influence ULK1 (a homolog of Atg1) protein expression. We used a non-small cell lung cancer cell line, Calu-1, to test the effect of TIA1 on ULK1 levels and autophagy. We found that the knockdown of TIA1 led to a reduction in protein levels of ULK1 in both the presence and absence of serum (Figures 4A and 4B) Crucially, stable knockdown of TIA1 using small interfering RNA (siRNA) had no effect on the transcript levels of ULK1, suggesting post-transcriptional regulation (Figure 4C). Next, we analyzed MAP1LC3B/LC3B-II levels using immunoblotting, which showed a consistent decrease post knockdown (KD) of TIA1 following serum starvation (Figures 4A and 4D). Treatment with bafilomycin A₁ blocks a late stage of autophagy and resulted in a substantial increase in ULK1 in the WT cells. In contrast, this treatment resulted in a significantly reduced accumulation of LC3B-II in TIA1 KD cells, indicating the reduction of autophagy in these cells as a result of lowered ULK1 levels (Figure 4D). Finally, we investigated whether TIA1 directly interacted with ULK1 mRNA. Indeed, TIA1 showed significant enrichment at the 3′ UTR of ULK1, similar to Pub1 (Figure 4E). Taken together, these results indicate that TIA1 plays a significant role in regulating ULK1 expression and autophagy, emphasizing the conserved nature of the regulation at the 3′ UTR of both ATG1 and ULK1 via Pub1 and TIA1, respectively.

Npl3 coordinates with Pub1 to export ATG1 mRNA and recruit polysomes for translation

Our investigations into the ATG1 mRNA interactome revealed Npl3 and Pub1 as post-transcriptional regulators. Previous high-throughput studies have shown that Npl3 interacts with Pub1.²⁵ This led us to ask whether Pub1 and Npl3 coordinate to enhance the translation of ATG1 during nitrogen starvation. To examine this, we performed co-immunoprecipitation of Npl3 tagged with PA and determined the interaction of Pub1-tagged with sfGFP using immunoblot. Consistently, Npl3 interacted with Pub1 during nitrogen-starvation conditions but not in the presence of nitrogen, suggesting that this interaction was specific to autophagy-inducing conditions (Figure 5A). Furthermore, this interaction was dependent on RNA; RNase treatment of the cell lysate prior to immunoprecipitation essentially eliminated the interaction between the two proteins. This finding suggests that Npl3 and Pub1 might coordinate to post-transcriptionally regulate the ATG1 transcript. Therefore, we tested whether the two proteins worked together in the polysome fraction. Toward this end, we tested the localization of Npl3 and Pub1 in the different fractions of free mRNPs, monosomes, and polysomes. Npl3 showed no enrichment in the monosome and polysome fractions (Figure S4A). In contrast, Pub1 interacted with both monosome and polysome fractions during nitrogen starvation, co-sedimenting with the cytosolic ribosome marker Rpl15 (Figure 5B). This finding suggests that Pub1 may have a more direct role in translational regulation of ATG1, while Npl3 may have an indirect role.

Considering that Npl3 did not interact with polysome components, we were interested to further dissect its role and localization in regulating ATG1 at the post-transcriptional level. In line with this, we found that following growth in nutrient-rich conditions and during nitrogen starvation, Npl3 was predominantly localized to the nucleus (Figure 5C). When we investigated the localization of Pub1 under the same conditions, we found that Pub1 was localized throughout the cell except in certain organelles such as the vacuole. Confocal imaging of Pub1-sfGFP along with a nuclear marker such as Nup116-mCherry showed that a subpopulation of Pub1 was present in the nucleus, suggesting that Pub1 acted as a nucleo-cytoplasmic shuttling protein while Npl3 was predominantly a nuclear protein (Figure S4B). Alternatively, we fractionated the cytoplasm and nucleus following nitrogen starvation and investigated the presence of Pub1 in the different fractions. We found that a small population of Pub1 was indeed found in the nucleus. Npl3, a nuclear protein, therefore interacted with the small population of Pub1 that was located in the nucleus.

To understand the importance of nuclear localization of Npl3, we used a point mutation that prevents nuclear shuttling. Previously, studies have shown that the point mutation E409K results in a protein predominantly localized to the cytosol.²⁶ This mutation is present in the region of Npl3 required for nuclear localization and results in a slower rate of import of Npl3 to the nucleus.²⁶ We constructed a strain containing this mutation and asked whether the E409K mutant showed defects in Atg1 expression. Indeed, Npl3^E409K had lowered Atg1 protein during nitrogen starvation but not ATG1 mRNA levels (Figures S4D–S4F). This defect resulted from the severe reduction in ATG1 mRNA binding exhibited by the E409K mutant (Figure 5D). This finding suggests that nuclear localization of Npl3 is required for nitrogen-starvation-dependent upregulation of Atg1 protein expression, likely via the interaction with Pub1. Npl3 interacts with the transcription machinery and loads transcripts onto mRNPs responsible for transcript export, underscoring its pivotal role in the nucleus.²⁷ These findings suggest a mechanism whereby the surveillance protein Npl3 recruits Pub1 onto the ATG1 transcript. Therefore, we asked whether the interaction of ATG1 mRNA and Pub1 is dependent on the presence of Npl3. RNA immunoprecipitation of Pub1 in WT and npl3Δ backgrounds showed that the interaction of Pub1 and the ATG1 3′ UTR was severely reduced in the absence of NPL3 (Figure 5E). Given that localization of Npl3 in the nucleus is critical for the Npl3-ATG1 interaction and Npl3 is required for Pub1-ATG1 interaction, we hypothesized that ATG1 loading onto Pub1 by Npl3 is required to export ATG1 transcript to the cytosol. Therefore, we asked whether pub1Δ cells also have an export defect of ATG1 mRNA. We found that the percentage of mRNA in the nuclear fraction was much higher in the absence of Pub1 compared to the WT (Figure 5F), thus phenocopying npl3Δ cells.

An obvious question was, does Pub1 link ATG1 mRNA export and translation? Pub1 interacts with polysomes, and several independent studies have shown that it also can interact with mRNA export proteins.²⁸ However, Pub1 is also a core stress granule component, with primary functions in the cytoplasm. Therefore, we wanted to determine all the interactions of Pub1 during nitrogen starvation and thus performed immunoprecipitation of Pub1-sfGFP followed by mass-spectrometry analysis to identify the factors that interact with Pub1 (Figure 5G). From this analysis, we confirmed the interaction of Pub1 with Npl3. Furthermore, we found that Pub1 interacted with several mRNA export proteins, along with Npl3. Notably, Pub1 interacted with Yra1 and Sem1, whose role in mRNA export from the nucleus has been long established.²⁹ We found that Pub1 bound to over 40 ribosome components, including large and small subunits, and interacted with translation factors such as Sui1 and Gcd1, involved in translation initiation, as well as Yef3, a translation elongation factor, thus suggesting that Pub1 indeed shuttles between the nucleus and the cytoplasm, linking mRNA export and translational machinery. Furthermore, Pub1 interacted with known RBPs that influence Atg1 protein levels, such as Dhh1, Pat1, Ded1, and Cdc33 (Data S3), confirming that Pub1 links the ATG1 mRNA-protein interactome to the translational machinery and ribosome components to aid in translation. These results also suggest that Pub1 mediates two spatially separated processes of nuclear mRNP loading and cytoplasmic translation, providing a unifying mechanism of how previously identified interactors of ATG1 help to deliver the transcript to translating ribosomes to enhance the Atg1 production needed to sustain starvation-dependent autophagy.

Limitations of the study

Our study primarily relies on biochemical methods including RNA immunoprecipitation and co-immunoprecipitation to capture protein-RNA interaction and protein-protein interaction, respectively. While this is a standard assay in the field, it fails to differentiate between direct and indirect interactions. Therefore, to provide better insight into these interactions, in vitro experiments with purified proteins and in vitro transcribed mRNA are required. Next, we employed polysome profiling to determine the translation status of the ATG1 transcript. The caveat with this experiment is that during nitrogen starvation the total number of polysomes is severely reduced, due to the block in global translation. Therefore, the yield of ribosomal fractions obtained is low, and further downstream analysis can further exacerbate technical artifacts. Additionally, this technique does not provide information regarding the region of the transcript that is translated. To circumvent these issues, a high-throughput approach such as ribosome profiling may be employed. We also provide insight into the nutrient-dependent role of Npl3 and Pub1; however, we did not identify an upstream cue such as a signaling protein or post-translational modification that drives the regulation. Further work will be needed to address this issue.

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Prof. Daniel J. Klionsky ([email protected]).

Materials availability

Plasmids and yeast strains will be made available upon request to the lead contact.

Data and code availability

• The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PRIDE: PXD059054.

• No new code was generated in this paper. Please contact the corresponding author for any other data requirements.

EXPERIMENTAL MODEL AND STUDY PARTICIPANTS

METHOD DETAILS

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistical analysis was done using GraphPad Prism from 2 to 3 independent biological replicates using either Student’s t test, one-way ANOVA or two-way ANOVA, corrected for multiple comparisons using Tukey’s test. For all figures, a p value < 0.05 was considered significant.